Arpita
Roy‡
a,
Subhendu
Dhibar‡
*b,
Kripasindhu
Karmakar
b,
Sangita
Some
b,
Sk Abdul
Hafiz
c,
Subham
Bhattacharjee
c,
Bidyut
Saha
*b and
Soumya Jyoti
Ray
*a
aDepartment of Physics, Indian Institute of Technology Patna, Bihar-801106, India. E-mail: ray@iitp.ac.in
bColloid Chemistry Laboratory, Department of Chemistry, The University of Burdwan, Golapbag, Burdwan-713104, West Bengal, India. E-mail: sdhibar@scholar.buruniv.ac.in; bsaha@chem.buruniv.ac.in; Tel: +91 7001575909 Tel: +91 9476341691
cDepartment of Chemistry, Kazi Nazrul University, Asansol-713303, West Bengal, India
First published on 28th February 2024
A well-organized, facile method for the development of a rapid supramolecular metallohydrogel of Zn(II)-ions, i.e. Zn@5AP, has been established using 5-amino-1-pentanol as a low molecular weight gelator (LMWG) in water at room temperature. The mechanical properties of the synthesized hydrogel were characterized through angular frequency and oscillatory stress-dependent rheological study. The self-healable nature of Zn@5AP was established through thixotropic analysis. Hierarchical microstructural features were characterized through field-emission scanning electron microscopy (FESEM) and tunneling electron microscopy (TEM) investigation. The EDS elemental mapping confirmed the primary chemical composition of the metallohydrogel. The potential metallohydrogel formation mechanism has been analyzed by FT-IR spectroscopic studies. Furthermore, a zinc(II) metallohydrogel (Zn@5AP)-based Schottky diode device in a lateral metal–semiconductor–metal geometry was fabricated to explore the charge transport behaviour. The resistive random access memory (RRAM) device based on zinc(II) metallohydrogel (Zn@5AP) exhibited bipolar resistive switching behavior at ambient temperature. The RRAM device offers exceptional switching endurance over 5000 consecutive switching cycles with a high ON/OFF ratio of ∼100. Due to the easy fabrication process, robust resistive switching behaviour, and enhanced stability of this system, these structures are suited for usage in non-volatile memory design, neuromorphic and in-memory computing, flexible electronics, and optoelectronics devices.
Hydrogels are an important class of supramolecular gel materials. The term hydrogel defines a three-dimensional network obtained from the immobilization of aqueous solution by LMWG molecules. The discovery and design of LMWGs with the ability to store aqueous solvent for the hydrogelation process are increasing day by day.12 Various LMWGs with good reputations in hydrogelation include fatty acids,13 steroids,14 amino acids,15 nucleic acids,16 peptides,17 and carbohydrates.18 Overall, hydrogels can be classified based on various properties, including the nature of associated groups, rheological and morphological structures, preparation method, constituting elements, non-covalent interactions, external stimuli, etc.19
Metal ions in metallohydrogels play an important role in the gelation process by scavenging the water solvent through molecular interactions with the gelator molecules.20 The metal ions can act functionally or simply as coordinating entities, as part of gelators or as auxiliaries. In the supramolecular hydrogel class, metallohydrogels are renowned for their extensive response to chemical and physical stimuli.21 Supramolecular metallohydrogels have become an important topic in the material science arena for their wide range of potential applications in drug delivery,22 artificial skin,23 sensing,24 tissue repair,25 semiconducting materials,26 dye absorption,27 optically active materials,28 electronic devices,29etc.
Modern science has recently been in high demand for high performance self-healing hydrogel materials for multiple purposes.30 Convenient metallohydrogel materials with rapid regaining of their structural integrity are in high demand for myriad applications in the biomedical field, such as for drug delivery.31 Materials with self-healing features have been developed by many researchers over the past few decades. The common type self-healing materials, i.e. polymers or elastomers,32 ceramics, fibre reinforced composites, cementitious materials, and metal-ion polymer systems33 are already established. Specifically, the sensitivity of self-healing hydrogels to temperature, chemical responsiveness, pH and enzymes can rapidly identify cell transport and drug delivery that can gratify the healing and tissue redevelopment of bio-systems.34 Besides the biological relevance, using self-healing hydrogel materials, some scientific groups are active in developing electronic devices,35 soft-robotics,36 3D/4D printing37 and load-bearing applications.38
For the last few years, our group have been actively working in the field of metallogel-based semiconducting devices and memory switching applications.39,40 However, self-healable metallohydrogels are still challenging for scientists. On the other hand, due to their superior memory properties, resistive random-access memory41–45 (RRAM) devices offer a wide range of applications in switching, non-volatile memory design, neuromorphic computing, etc. In RRAM devices the switching can occur between the high and low resistance states.47,48 The mechanism can be explained by different physical processes, such as charge carrier trapping and vacancy migration, electrochemical migration, etc. Due to its compatibility with CMOS architecture, simple structure, good manufacturability, low cost, low power consumption, high speed, long durability, and dependability, it is an advantageous technology for next-generation memory design and in-memory computing where computing and storage can occur at the same place to offer distinct advantages over the current Von-Neumann based architecture in terms of processing speed and time lag in the hardware. The switching behavior of oxide-based RRAM devices49–54 has been demonstrated in our recent works. However, researchers are constantly looking for alternative materials and devices to improve their switching behavior and metallohydrogels can be a useful candidate due to their semiconducting nature and device integration capacity. Moreover, metallohydrogel-based RRAM structures46 can be developed on flexible substrates to design flexible electronic and optoelectronic devices useful for memory, sensing and optical detection.
Memristors have a large application in memory, logic operations and neural synaptic networks due to their non-volatility and nonlinearity. Our work presents a concise overview of memristor-based logic circuits and analyses their role in memory applications. Furthermore, this work explores the possibilities of utilizing memristors for implementing logic in memristor arrays. Memristive devices with in-memory computing technology differ from conventional computing systems, where logical operations and data storage are different by presenting greater potential in artificial intelligence applications. Therefore, research on memristor-based logic circuits opens new possibilities and methodologies for designing innovative logic architectures. These memristor-based logic operations are categorized as an exploration of novel logic circuits.
Through the present work, we attempted to establish an eco-friendly method to achieve a rapid self-healable zinc(II) and 5-amino-1-pentanol-based metallohydrogel under ambient reaction conditions at room temperature using water as a solvent (Fig. 1). The available amine and hydroxyl groups in the gelator are responsible for rapid gelation via non-covalent interactions with Zn(II)-ions in the presence of water molecules and the resulting gel network acts as an ideal soft scaffold. Based on the synthesized Zn(II)-metallohydrogel (Zn@5AP) mediated metal–semiconductor (MS) junctions, we have successfully developed non-volatile resistive random access memory (RRAM) devices. Our approach to creating a versatile, useful soft gel scaffold may develop the field of memory devices based on science and technology for applications in neuromorphic computing and data-driven applications like the Internet of Things (IoT), 5G connectivity, etc.
The rheological investigation was performed utilizing a 20 mm SS parallel plate geometry and a Peltier plate temperature system set to 25 °C on a DHR-2 stress-controlled rheometer, supplied by TA Instruments.
FESEM pictures were collected by a Carl Zeiss SUPRA 55VP instrument.
TEM images were conducted in an aberration corrected FEI Titan Themis operating at 300 kV.
The ZEISS EVO 18 microscope was used to conduct elemental mapping and EDX tests.
A JASCO FTIR 4700 spectrometer was used to collect the FTIR data.
The current–voltage (I–V) measurements of our synthesized metallohydrogel material-based devices were performed at room temperature using a Keithley 2400 sourcemeter interfaced through Labview.
However, the storage modulus (G′) crosses loss modulus (G′′) when the strain reaches ∼0.1%, demonstrating gel-to-sol transition. Finally, to discern the self-healing behavior, a thixotropy test was performed (Fig. 3c). Interestingly, the soft material remains in the gel state when the strain is ∼0.01%, but it turns into a sol immediately upon sudden increase of the strain to ∼100%. Upon removal of the high strain, the soft material turns into a viscoelastic gel state almost immediately. Such successive low/high strain cycles were repeated, which clearly showed the reversible transition of the soft material from gel-to-sol state, which in turn, demonstrated the self-healing behaviour of the metallohydrogel.
The column of pure Zn@5AP metallohydrogel for testing the self-healing property was prepared following the synthesized method (Fig. 4a). The column was cut into three pieces with dimensions of approximately ∼1–2 cm (Fig. 4a). Block pieces of Zn@5AP were brought back into contact with slight pressure and fully welded to each other in <5 min (Fig. 4a). The cured monolith exhibited a strong edge capable of free standing without any external support (Fig. 4b). The stability of the monoliths after curing against gravity is also revealed in Fig. 4c and d. Furthermore to investigate the molding nature of the Zn@5AP metallohydrogel, it was formed into the shape of a toy car and heart without losing any fluid and maintained its stability without external support, demonstrating its extraordinary moldable properties (Fig. 4e and f). In addition, a sketch was drawn to show the use of the Zn@5AP metallohydrogel as an ink (Fig. 4g). Zn@5AP metallohydrogel was also used to prepare nodule-shaped patterns using a surgical syringe (Fig. 4h), demonstrating that Zn@5AP metallohydrogel can be used for injection.
Self-assembly of Zn(II) ions and 5-amino-1-pentanol is the outcome of hydrogelation through various supramolecular interactions like hydrogen bonding patterns, appropriate metal–ligand coordination, electrostatic connections, and hydrophobic interactions. These microstructures were part of the layered structure of the Zn@5AP metallohydrogel and had an average diameter of about ∼1 μm. High resolution TEM images show the hemi-spherical shape of the Zn@5AP metallohydrogel samples (Fig. 5c and d). EDS elemental mapping in Fig. 5e–i confirms the presence of Zn, C, O and N elements of Zn(NO3)2·6H2O, the 5-amino-1-pentanol gelator and the water solvent, accountable in Zn@5AP metallohydrogel networks.
(αhν)n = A(hν − Eg) | (1) |
In electron transition processes, the exponent “n” functions as a dependent constant. “A” refers to a constant with the value 1. Using n = 2, the direct optical band gap was computed. We determined the direct optical band gap (Eg), which is assumed to be 4.55 eV, by expanding the linear region of the plot (αhν)2versus hν (Fig. 7) to α = 0 absorption.
In order to design RRAM (resistive random access memory) devices based on the Zn@5AP metallohydrogel, vertically placed layers in a sandwiched structure of ITO/Zn@5AP/Cu (called device 2) and Cu/Zn@5AP/Cu were prepared (named device 3). The bottom electrodes were cleaned using an ultrasonicator in both designs, which were followed by the deposition of the Zn@5AP metallogel on the substrate and the addition of the top electrode on the samples. The thickness of Zn@5AP is around 200 nm and the dimension of the device is 1 cm × 1 cm. The device structures are given in Table 1.
Device no. | Device structure | Dimensions of the device |
---|---|---|
Device 1 | ITO/Zn@5AP/ITO | 1 cm × 1 cm |
Device 2 | ITO/Zn@5AP/Cu | 1 cm × 1 cm |
Device 3 | Cu/Zn@5AP/Cu | 1 cm × 1 cm |
In order to get a detailed understanding of the semiconducting nature of the Zn@5AP metallohydrogel in the thin film geometry, its charge transport properties were measured in a two-probe configuration. There is no discernible conduction in the voltage region (−5 V to +5 V), where the I–V curve of device 1 is shown in Fig. 8 on a linear scale. The current was found to increase rapidly with an increase in voltage in both the positive and negative polarities. The main diode parameters were then extracted from the non-linear Schottky diode's I–V curve using the thermionic emission theory (TE Theory) after analysing the I–V data as suggested by Cheung.47 We note that the semiconductor with SBs at both contacts can be modeled with two back-to-back Schottky diodes separated by a series resistance. When a sufficiently high external voltage is applied, whether a positive or negative, one Schottky junction is forward-biased while the other one is reverse-biased. The reverse saturation current of the reverse-biased diode always limits the current.
Based on thermionic emission, the current at the two contacts can be written as,
(2) |
(3) |
(4) |
(5) |
(6) |
(7) |
(8) |
(9) |
(10) |
(11) |
(12) |
(13) |
(14) |
H(I) = IRS + ηϕB | (15) |
The series resistance, ideality factor, and barrier potential height were calculated using eqn (13)–(15). These formulas were produced by simply deriving from Cheung's theory. We plotted the dV/d(lnI) vs. I graph and H vs. I graph, as shown in Fig. 9, in order to determine the diode parameters of device 1. We determined the barrier height from the intercept of the H vs. I graph and the ideality factor (η) from the intercept of the dV/d(lnI) vs. I graph. The ideality factor (η) for our diodes was determined to be 25.13, which is greater than the ideal value of 1.1. The series resistance at the interface, the presence of interface states, and the presence of inhomogeneities in the Schottky barrier55 could all contribute to the divergence of the expected behaviour.
A barrier height (ϕB) of 0.03 eV was determined for device 1. The fabricated diode structure clearly has a lower barrier potential height and a greater ideality factor and these two characteristics are essential for designing a Schottky diode. We computed the slope of the dV/d(lnI) vs. I graph as well as the H vs. I graph in order to obtain the value of the series resistance, which is 5963.04 Ω. After taking into consideration all of the measured properties, it can be stated that such diode layouts can be useful for semiconductor electronics design. We have also measured I–V characteristics at different low temperatures from 100 K to 200 K, as shown in Fig. S4 (ESI†). I–V measurements are also used to examine the resistive switching behaviour of the Zn@5AP metallohydrogel-based heterostructure (see Fig. 10 (right: inset)) where ITO acts as the bottom electrode and Cu acts as the top electrode in device 2. To avoid any leakage contribution to the I–V measurements, the compliance current (CC) is set at 100 mA before the commencement of any tests. The I–V trajectory of device 2 is displayed on a linear scale in Fig. 10. In this instance, the I–V measurements are performed in the following sequence: 0 V → 5 V → −5 V → 0 V. The applied voltage is shown by the arrows. The I–V loop displays complete hysteresis, which is an indication of memristive behaviour. The current across the low-voltage zone initially rises linearly when the voltage is close to 0 V (point A to point B). Non-linearity in the current–voltage graph is seen under a specific voltage (2.829 V), and as the applied voltage increases, the current increases abruptly. The device is now transitioning from a higher resistance state (HRS) to a lower resistance state (LRS), according to the SET voltage (VSET = 2.829 V). At point C, the current increases. Then, even if the voltage is switched from point C to point D, the device maintains its ON state. After the voltage drops, the current starts to rapidly decrease at point E, and the device turns into the OFF state. After cycling in the opposite direction, the current keeps increasing and achieves an LRS at a value of −2.43 V (point E).
At a RESET voltage (VRESET) of −4.8 V, the device returns to the HRS. The system changes from LRS to HRS using the RESET method, which is denoted by the arrow between points F and G. This is a crucial indicator of the sample's bipolar resistive switching behaviour because a negative voltage is required to return the sample to its prior resistance state. The I–V properties of device 2 are displayed on a semi-logarithmic scale in Fig. 10 (left: inset). The increase and decrease in current caused by the formation and rupture of conductive filament-type structures tries to switch the device between the ON and OFF states. We have also measured the complete I–V curves of device 2 for consecutive cycles and observed some variability in the first few cycles (∼20) as shown in Fig. S1 (ESI†) which gets stabilized. Later, this mechanism is described.
In order to better understand the conduction mechanism and the charge transport mechanism of device 2, we have fitted the I–V curve in the logarithmic scale (Fig. 11a) throughout the SET process. In this instance, we have come to the conclusion that the current exhibits ohmic conduction and changes linearly in the voltage range of 0 V to 3 V with a slope of m = 1. But in the higher voltage region, the current follows space charge limited conduction behaviour with a slope of m = 2.16. It is confirmed that it is followed by space charge conduction behaviour when the slope is around 2. Similar behaviour has also been exposed by the RESET process (Fig. 11b). We have noticed that it exhibits ohmic behaviour with a slope of m = 1.2 in the lower voltage region between 0 V and 3 V.
In order to observe the resistive switching characteristics of device 3, Cu is used as both the top and bottom electrodes in device 3. In the presence of the SET and RESET operations (VSET = 4.98 V and VRESET = −4.58 V), the I–V curve exhibits memristive behaviour over full voltage cycling with discrete LRS and HRS states, as shown in Fig. 12. This behaviour resembles the I–V curve of device 2 in the same way. We have also measured the complete I–V curves of device 3 for consecutive cycles and observed some variability in the first few cycles (∼20) as shown in Fig. S2 (ESI†) which gets stabilized later. For comparing device 3 with device 2, we have observed that there is some variation in the I–V curve in the complete cyclic response, which is shown in Fig. S3 (ESI†).
For a better understanding of the conduction mechanism and the charge transport process of device 3, we have also fitted the I–V curve in the logarithmic scale in the SET process (Fig. 12 (right: inset)). We have observed that the current follows the ohmic nature of conduction and varies linearly with a slope of m = 1.76.
We also plotted the I–V curve using a logarithmic scale during the RESET procedure (Fig. 13). Here, we have seen that it exhibits ohmic behaviour with a slope of 1.13 in the voltage range between 0 V and 4 V.
Fig. 13 I–V curve of the glass/Cu/Zn@5AP/Cu based device on a logarithmic scale in the RESET process. |
To better understand the consistency of the switching process, we also performed an endurance test for 5000 switching cycles consecutively at room temperature for both the devices (2 and 3), as shown in Fig. 14(a) and (b). The switching procedure for device 2 and 3 is reliable for up to 5000 switching cycles. The endurance test shows that the switching behaviour is robust because the average ON/OFF ratio is around 100 for both the devices. This implies that this device can continue to function as intended in terms of memory response over an extended period without experiencing any degradation, which is helpful for real-world uses in memory circuit design at a reduced cost of production.
Fig. 14 (a) Endurance test of the glass/ITO/Zn@5AP/Cu based device; (b) endurance test of the glass/Cu/Zn@5AP/Cu based device. |
We have also performed retention tests for up to 1000 s at room temperature for both the devices (2 and 3), as shown in Fig. 15(a) and (b). The switching procedure for devices 2 and 3 is reliable for up to 103 s. We have also observed that for both the devices (2 and 3) the ON/OFF ratio is around 85. From the retention test, we can conclude that these devices can store data up to 103 s without any degradation.
Fig. 15 (a) Retention test of the glass/ITO/Zn@5AP/Cu based device; (b) retention test of the glass/Cu/Zn@5AP/Cu based device. |
A comparison of ION/IOFF observed in other materials at room temperature (shown in Table 2).
Numerous processes, such as the formation of the Schottky barrier with electrochemical migration, redox reactions, valence change memory, etc. can be used to explain the physical origin of the switching process. According to our study, the valence change memory (VCM) and electrochemical metallization (ECM) process are impacted by the mobility of oxygen defects and metal cations. Here, it is shown that metal ions and oxygen vacancies both have a substantial impact on the change in resistance (Fig. 16). For device 2, conducting filaments form in the semiconducting layers and as a result there is the migration of Cu ions, Zn ions, and oxygen vacancies. We can explain the transition from the HRS to LRS state by rupturing and regenerating the Cu filaments that are integrated into the semiconducting layers. In this study, the resistive switching behaviour of device 2 is mainly caused by the migration of Cu ions and oxygen vacancies. We already know that copper ions can move in the direction of an applied electric field and ionize in the presence of an electric field to generate copper ions with the formula Cu → Cu2+ + e−. When a positive voltage is applied, Cu2+ ions, Zn ions, and oxygen vacancies travel towards the intermediate layer where they are transformed into metallic Cu. After the SET process is complete and the device switches from HRS to LRS, the conductivity of this layer will rise and the concentration of Cu ions, Zn ions, and oxygen vacancies will approach the bottom electrode. Until a sufficient voltage with the opposite polarity is applied to electrochemically dissolve the Cu filaments and oxygen vacancies for the RESET operation, the device stays in the LRS state. The device enters the HRS when negative voltage is applied, and its conductivity also drops at the same moment. At the end, Cu2+, Zn, and oxygen vacancy ions return to the upper electrode. In this way, the resistive switching behaviour of device 3 can be described by the migration of Cu2+ ions, Zn ions, and oxygen vacancies. We have also described the conduction filament model from TEM and EDAX analysis, as shown in Fig. S5 (ESI†) (Table 3).
RRAM device | Endurance (cycles) | Retention (s) | ON/OFF ratio | Power consumption (nW) | Ref. |
---|---|---|---|---|---|
Al2O3/ZnO-based RRAM | 104 | 104 | 105 | 1.4 × 104 | 56 |
Pt/AlOx/ZnO/Ti | 103 | 104 | 102 | 0.6 | 57 |
Pt/HfO2/TiOx/Pt | 100 | 104 | 102 | 1.12 × 103 | 58 |
Graphene/HfO2/TiN | 103 | 104 | 102 | 4.6 × 102 | 59 |
hhBN/AlOx/TiOx/Ito | 100 | 104 | 102 | 2 × 103 | 60 |
Cu/Zn@5AP/ITO | 5000 | 103 | 102 | 4.72 × 103 | This work |
Memory devices in memory systems are organized in arrays, and one common architecture is the cross-point architecture. In this configuration, our sample is placed at the crossing point of two perpendicular metal lines, as shown in Fig. 17. In our research work, we have shown how the metallogel-based RRAM device in a crossbar array works in memory computing using logic gate operation. In-memory computing can also be accomplished by employing fundamental electrical circuit principles, such as Kirchhoff's law and Ohm's law. This method is particularly well-suited for analog crossbar arrays. Here, in this work we have prepared a 2 × 2 cross bar device based on a Cu/Zn@5AP/Cu structure where Cu acts as both the top and bottom electrode as shown in Fig. 17 and the four RRAM devices are denoted by A, B, C, and D.
Now, for logic gate operations we have used two RRAM devices (A and B). For the OR logic gate, if memristors A and B are both logical “0” that means no voltage is applied, and then the output voltage is 0.143 V which is considered as the “0” state. When we applied 5 V at input A and 0 V at input B, then the output voltage is 4.96 V which is considered as the logical “1” state. Similarly, when we applied 0 V at input A and 5 V at input B, then the output voltage is 4.95 V which is also considered as the logical “1” state. When both memristors A and B are both in the logical “1” state, that means when we applied 5 V at both the terminals, then the output voltage is also 4.95 V as the logical “1” state. The truth table of the OR logic gate is shown in Table 4.
Voltage at input A (V) | Voltage at input B (V) | Output voltage (V) | Logical state |
---|---|---|---|
0 | 0 | 0.14 | 0 |
5 | 0 | 4.96 | 1 |
0 | 5 | 4.95 | 1 |
5 | 5 | 4.95 | 1 |
Similarly, we have designed a NOT gate logic circuit (shown in Fig. 17) using device C and it also satisfied a NOT gate truth table, which is shown in Table 5.
Input voltage | Output voltage | Logical state |
---|---|---|
0 V | V | 1 |
5 V | 0.11 V | 0 |
The current structure can be extended further with a larger size of cross-point arrays to perform advanced logic and computing operations, which can act as a central part for in-memory computing where the computation and information storage are done at the same circuit level, as demonstrated here. In this way, memristor based logic gate circuits using crossbar arrays will help us to explore different engineering methodologies that depend on in-memory computing principles.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3ma00765k |
‡ AR and SD should be treated as joint first authors |
This journal is © The Royal Society of Chemistry 2024 |