Science and Technology of Advanced Materials

ISSN: 1468-6996 (Print) 1878-5514 (Online) Journal homepage: www.tandfonline.com/journals/tsta20

Decade of 2D-materials-based RRAM devices: a

review

Muhammad Muqeet Rehman, Hafiz Mohammad Mutee Ur Rehman, Jahan

Zeb Gul, Woo Young Kim, Khasan S Karimov & Nisar Ahmed

To cite this article: Muhammad Muqeet Rehman, Hafiz Mohammad Mutee Ur Rehman, Jahan
Zeb Gul, Woo Young Kim, Khasan S Karimov & Nisar Ahmed (2020) Decade of 2D-materials-
based RRAM devices: a review, Science and Technology of Advanced Materials, 21:1, 147-186,
DOI: 10.1080/14686996.2020.1730236

To link to this article: https://doi.org/10.1080/14686996.2020.1730236

© 2020 The Author(s). Published by National

Institute for Materials Science in partnership
with Taylor & Francis Group.

Published online: 18 Mar 2020.

Submit your article to this journal

Article views: 17247

View related articles

View Crossmark data

Citing articles: 57 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tsta20
SCIENCE AND TECHNOLOGY OF ADVANCED MATERIALS
2020, VOL. 21, NO. 1, 147–186
https://doi.org/10.1080/14686996.2020.1730236

Decade of 2D-materials-based RRAM devices: a review

Muhammad Muqeet Rehmana, Hafiz Mohammad Mutee Ur Rehmanc, Jahan Zeb Gulb, Woo Young Kimc,
Khasan S Karimova and Nisar Ahmeda
a
Faculty of Electrical Engineering, Ghulam Ishaq Khan Institute of Engineering Sciences and Technology, Topi, Pakistan;
b
Department of Mechatronics & Biomedical Engineering, AIR University, Islamabad, Pakistan;
c
Faculty of Electronic Engineering, Jeju National University, Jeju, South Korea

ABSTRACT ARTICLE HISTORY

Two dimensional (2D) materials have offered unique electrical, chemical, mechanical and Received 2 November 2019
physical properties over the past decade owing to their ultrathin, flexible, and multilayer Revised 12 February 2020
structure. These layered materials are being used in numerous electronic devices for various Accepted 12 February 2020
applications, and this review will specifically focus on the resistive random access memories KEYWORDS
(RRAMs) based on 2D materials and their nanocomposites. This study presents the device 2D materials; resistive
structures, conduction mechanisms, resistive switching properties, fabrication technologies, switching; nonvolatile;
challenges and future aspects of 2D-materials-based RRAMs. Graphene, derivatives of gra- bipolar & unipolar; RRAMs;
phene and MoS2 have been the major contributors among 2D materials for the application of fabrication technology;
RRAMs; however, other members of this family such as hBN, MoSe2, WS2 and WSe2 have also planar & sandwiched
been inspected more recently as the functional materials of nonvolatile RRAM devices. structure
Conduction in these devices is usually dominated by either the penetration of metallic ions CLASSIFICATION
or migration of intrinsic species. Most prominent advantages offered by RRAM devices based 105 Low-Dimension (1D/2D)
on 2D materials include fast switching speed (<10 ns), less power losses (10 pJ), lower threshold materials; 201 Electronics /
voltage (<1 V) long retention time (>10 years), high electrical endurance (>108 voltage cycles) Semiconductor / TCOs; 306
and extended mechanical robustness (500 bending cycles). Resistive switching properties of 2D Thin film / Coatings; 503
materials have been further enhanced by blending them with metallic nanoparticles, organic TEM; STEM; SEM; Memory
polymers and inorganic semiconductors in various forms. devices; RRAMs

1. Introduction
difficulty of charge retention in dwindling dimensions
Flash memory devices based on silicon have ruled the [5]. Typical memory devices have served the needs of
electronic industry due to their successful and distin- data storage for several years but are no longer
guished nonvolatile memory (NVM) characteristics a solution to ever increasing demand for high-
[1–3]. Nevertheless, existing NVM devices including density memory storage. Information storage technol-
flash memories face a few severe problems in their ogy based on silicon is facing physical and theoretical
operation of charge storage such as slow programming limitations. Therefore, the challenge of exploring and
speed, relatively high operating voltage, and poor elec- developing new techniques and devices exist to over-
trical endurance [4]. Moreover, their endless shrinking come data storage deficiency. A comprehensive study
(to enhance the data storage capacity) is anticipated in this regard for the various available advanced mem-
soon to reach its basic scaling limit mainly due to the ory technologies has been reported [6].

CONTACT Muhammad Muqeet Rehman muqeetmughal88@gmail.com Faculty of Electrical Engineering, Ghulam Ishaq Khan Institute of
Engineering Sciences and Technology, Topi, KPK 23460, Pakistan
This article has been corrected with minor changes. These changes do not impact the academic content of the article.
© 2020 The Author(s). Published by National Institute for Materials Science in partnership with Taylor & Francis Group.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits
unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Sci. Technol. Adv. Mater. 21 (2020) 148 M.M. REHMAN et al.

As a substitute to the typical silicon-based memory, Although TMO-based RRAM devices have exhibited
resistive random access memory (RRAM) devices have the desired properties, the biggest challenge remains in
emerged as the potential candidate for the future flexible realizing a functional memory material having all
NVM device owing to their standout features such as desired attributes in it. For example, the most critical
simple structure [7], low power consumption [8], high trade-offs are power–speed, speed–retention, and
scalability [9], low cost (≈0.019 $ GB–1) [6], enormous endurance–retention which are not offered by a single
data storage (5.62 × 1010 bits cm–2) [6], highly desirable TMO-based RRAM device. TMOs based RRAM devices
multi-bit data storage/unit cell, compatibility with CMOS are mostly influenced by surface effects, including sur-
technology [10,11], and compatibility with fabrication face roughness [36], chemisorption/photodesorption
techniques [12,13]. Intermediary resistive states can also [37], and surface band bending [38]. For example, the
be obtained that give rise to added functionalities such as electrical properties of TMOs based RRAM devices can
neuromorphic processing, filters, logic, arithmetic and be affected by the chemisorption of O2 molecules. Its
multibit storage. Owing to these promising attributes, electrical conductivity decreases due to chemisorbed
47 billion USD has been invested in these memory oxygen ions. Furthermore, the existing functional mate-
devices in 2016 [6]. Furthermore, RRAM devices offer rials being used in RRAMs are facing the bottleneck of
flexibility unlike hard disks and flash drives that seems to scalability. Therefore, search for exploring the new func-
be the future of electronic industry. tional materials for RRAM devices remains an open and
NAND Flash is the most used memory these days with important research area for realizing the development of
a major disadvantage of current leakage when it reaches commercially available flexible and wearable NVM
the nanometer scale to achieve high density, mainly due device of the next generation. 2D materials such as
to the integration of capacitor in it. Current leakage can graphene, MoS2, hBN, WS2, MoTe2 and MoSe2
result in loss of information [14] therefore, resistive [39–56] have already been introduced as the active
memory devices are superior to capacitive memory layer sandwiched between the two conductive electrodes
devices due to high scalability without compromising to form RRAM devices on flexible and rigid substrates.
data storage. Some of the most remarkable RRAM Graphene was the first 2D material that was used in
devices reported till now include the RRAMs with electronic device applications followed by its other
a superfast switching speed (300 ps) [15–18], extremely family members like transition metal dichalcogenides
low power losses (0.1 pJ) [15,16], long electrical endur- (TMDs), hBN and MXenes. To date, graphene is the
ance (1012 voltage sweeps) [19–24], and long data reten- leading 2D material due to its electrical and mechanical
tion (several years) [25]. RRAM devices can be characteristics required for RRAMs and its ability to be
distinguished on the basis of their switching mechanism used as a highly conductive electrode material. One of
into either distributed/local switching or bipolar/unipolar the important characteristics of RRAM devices is their
switching [26]. switching ratio ION/IOFF, which can be as high as 109
Search for the right functional material for the for 2D-materials-based RRAM devices [57] and is ~106
RRAMs is a significant research area owing to their for TMO-based RRAM [58]. Furthermore, the first
attractive advantages. The most intensively used TMO-based RRAM device was fabricated around
RRAM devices to date with high reliability are based 60 years back in 1967 and compared to them 2D-
on insulating transition metal oxides (TMOs) such as materials-based RRAM devices have already began to
Al2O3 [10,16,27], TaOx [28], and TiO2 [29–32]. compete with them in only a single decade which is an
A connection is formed between the two electrodes encouraging sign for the researcher working on these
of these TMO-based memory devices in the form of materials. These 2D materials have unique chemical,
a conductive filament (CF) on the application of exter- electrical, thermal, mechanical, physical, magnetic, and
nal electric field. Switching between the high resis- optical properties due to their adjustable number of
tance state (HRS) and low resistance state (LRS) layers that makes a strong case for them to be further
occur due to the formation and dissolution of nano- explored as the functional layer of RRAMs. The stand-
sized filaments. TMO-based RRAMs are further clas- out feature of 2D materials-based RRAMs is their high
sified into two major categories of valence change flexibility, e.g. the Young’s modulus and yield stress of
memory (VCM) and electrochemical metallization graphene are 335 N.m−1 and 42 ± 4 N.m−1 respectively
(ECM) based on their physical conduction mechanism while the bending modulus of MoS2 is 9.61 eV [59].
[33]. Conduction in VCM is due to the phenomena of These values of most commonly used 2D materials for
oxygen vacancies movement within the TMO active RRAM devices suggest that they will offer great flexibil-
layer while metallic ions from the two electrodes are ity and mechanical endurance. Another important char-
responsible for causing switching in ECM devices [34]. acteristic of 2D-materials-based RRAMs is that they can
TaOx based memory devices have in general shown withstand extreme temperature conditions without
highest stability against the applied voltage sweeps compromising on their memory characteristics. 2D-
whereas the fastest switching speed and lowest energy materials-based RRAMs offer highly scalable memory
losses are the attributes of HfOx based RRAMs [35]. cells with fast switching speed and low power
Sci. Technol. Adv. Mater. 21 (2020) 149 M.M. REHMAN et al.

consumption. The switching voltages (SET and RESET) electrodes made of metals owing to the contact resis-
and switching ratio of such devices are also tunable. The tance offered by them but the power losses in graphene
detailed discussion about the 2D-materials-based electrodes based RRAM devices are low due to its high
RRAMs has been provided in the later sections of this conductivity and its role as a tunneling barrier.
review paper. Thirdly, significant contribution offered by graphene
In this review, we have summarized the resistive electrodes is high scalability due to their ultrathin
switching properties of RRAM devices based on 2D nature. Moreover, using graphene as an electrode
materials and their composites, recent advancements material offers the advantage of mechanical flexibility
in their performance, fabrication methods, their due to which next-generation nonvolatile RRAM
potential as the future NVM flexible RRAM and chal- devices can be used in the application of flexible
lenges faced in realizing its commercialization. electronics.
Furthermore, research done by each research group Keeping in view all the above-mentioned prospects
has been discussed in detail and their significant con- of graphene, Yu et al. [60] published a significant article
tributions have been highlighted to show their impor- in which they explored the resistive switching effect in
tance in the advancement of this topic. The trends of the flexible and transparent single-walled carbon nano-
publishing research papers by various research groups tube (SWCNT) with a distinction that graphene was
in the last decade of 2D-materials-based RRAM used as the electrode material. The device was treated
devices around the world shows their deep interest with ozone to develop a bonding between oxygen atoms
and future prospects of these devices as illustrated in and the surface of graphene that acted as trap sites for
Figure 1(a,b) while Figure 1(c) shows the data of a few charge carriers. The graphene electrodes were decorated
RRAMs with quickest switching speeds including 2D with oxygen because it did not decrease the transmit-
materials. tance of electrode as compared to Au and Al nanopar-
ticles that caused a considerable decrease in the
transmittance of graphene by 11% and 25%, respec-
2. Role of graphene and its derivatives in tively. This nonvolatile RRAM device based on gra-
RRAM devices phene electrode displayed a high mobility and fast
switching speed of ~44 cm2 V−1 s−1, and 100 ns, respec-
2.1. Role of graphene in RRAM devices
tively, with promising mechanical properties (103 bend-
2.1.1. Graphene as electrode in RRAM devices ing cycles) owing to the flexibility of graphene
Formation and rupture of filaments is one of the most electrodes as compared to other rigid metallic electro-
proposal conduction mechanism in RRAM devices. des like Ag, Au, Al and Pt. The electrical endurance,
The main problem faced by researchers was to verify retention time and optical transmittance of this RRAM
this mechanism through advanced technology like device were 500 voltage sweeps, 103 s and 83.8%, respec-
transmission electron microscopy (TEM). Graphene- tively. A photograph of the transparent graphene–car-
based electrodes resolve this problem owing to theirs bon nanotube (CNT) integrated circuit array on
high transparency and small thickness, hence allowing a polyethylene terephthalate (PET) film, along with an
the researchers to view the formation and rupture of optical image of a monolayer graphene–CNT transistor,
filaments within the active layer of their devices. is shown in Figure 2(a) while the entire fabrication
Secondly, promising property associated with the process of this RRAM device is illustrated in Figure 2
RRAM devices is its low power consumption that is (b). Ji et al. [61] reported another important research
often compromised due to the use of traditional work on the resistive switching properties of organic

Figure 1. (a) A bar graph illustrating the quantity of published articles on the topic of 2D-materials-based RRAM devices on a yearly
basis. (b) A Pi chart showing the ratio of graphene, MoS2 and other 2D-materials-based research articles. (c) A bar chart illustrating
the high switching speeds of RRAMs based on different functional materials including 2D materials.
Sci. Technol. Adv. Mater. 21 (2020) 150 M.M. REHMAN et al.

Figure 2. (a) Photograph of the transparent graphene–CNT integrated circuit array (16 × 18) with schematic of the graphene
electrode (source, drain and gate) and SWCNT network channel on the PET film. Optical image of a monolayer graphene–CNT
transistor (5 channels), in which the dark square rectangle between the source and drain indicates the oxide layer, and Sweeping
characteristics of the oxygen-decorated graphene/SWCNT memory device [60]. (b) Fabrication process of flexible organic memory
devices with transparent graphene top electrodes. (c) I–V characteristics of an MLG/PI:PCBM/Al flexible organic memory device
under different bending conditions and ON/OFF ratio of a flat memory device [61].

memory device fabricated on a flexible PET substrate techniques. A schematic diagram along with the I–V
with graphene electrodes. It was an 8 × 8 bar array with characteristics of the proposed device are illustrated in
transparent multilayer graphene electrodes. Unlike Yu Figure 3(a). Park et al. did not use the two important
et al. who used graphene as the bottom electrode for features of graphene electrodes, i.e. transparency and
their RRAM device, Ji et al. used multilayer graphene as flexibility in their proposed RRAM device based on
the top electrode for their organic memory device. This silicon oxide; however, Yao et al. [63] explored both
RRAM device exhibited a high switching ratio of 106 the above-mentioned properties of graphene electro-
and WORM behavior. The mechanical robustness (104 des in their proposed RRAM device based on silicon
bending cycles) displayed by this device was 10 orders oxide on a flexible substrate. They also studied the
more than Yu et al. with a good cell to cell uniformity effect of device size on the conduction mechanism
and without any prominent decay in device perfor- for high-density memory applications. The obtained
mance. The electrical characteristics of this device are results of Yao et al. were superior to those reported
shown in Figure 2(c). earlier by Park et al. as the electrical endurance and
Owing to the increasing interest in RRAM devices switching ratio of their memory device were 300 vol-
and potential of graphene to act as the electrode mate- tage sweeps and 105 respectively. A schematic diagram
rial in RRAM devices, Park et al. [62] carried forward and the optical image of this device are shown in
this work and published an article on the resistive Figure 3(b).
switching behavior of silicon oxide with graphene Lai et al. [64] developed a flexible, transferable and
electrodes. This device displayed unipolar resistive sticker type RRAM device with graphene electrodes
switching behavior that was independent of applied that could also be developed on arbitrary substrates
voltage polarity with average memory characteristics with high performance. They used PMMA polymer
of 102 switching ratio and electrical endurance of 80 for the dual purposes of insulating active layer of
voltage sweeps. However, the distinguishing factor of RRAM device along with acting as one of the active
this study was the high density of data that could be layers itself. Poly(3-butylthiophene) (P3BT) was the
stored in this RRAM, i.e. up to ~0.5 terabits per square major active material that was used in this device to
inch. The device was fabricated without using expen- store charge. They showed that chemical vapor
sive and complex pattern transferring or lithography deposition (CVD)-grown graphene electrodes were
Sci. Technol. Adv. Mater. 21 (2020) 151 M.M. REHMAN et al.

Figure 3. (a) Self-assembled silicon oxide memristors. Lamella-forming and sphere-forming block copolymers were used for
preparing thin film and nanodot memristors, respectively. Current (I)−voltage (V) measurement results for the metal/oxide/metal
samples and reliability – Pt/SiOx nanodot/Pt [62]. (b) The gate-controlled BLG-based RRAM. The device structure of the GC-GRRAM
and the main process flow of the GC-GRRAM [70]. (c) Structure of graphene-based and Pt-based RRAM in a vertical 3D cross-point
architecture. An illustration of graphene-based RRAM in a vertical cross-point architecture. The RRAM cells are formed at the
intersections of the TiN pillar electrode and the graphene plane electrode. The resistive switching HfOx layer surrounds the TiN
pillar electrode and is also in contact with the graphene plane electrode. A schematic cross-section of the graphene-based RRAM
with I–V curve [71].

very useful for 3D stacking, vertical integration and Owing to the promising attributes of graphene and
fabrication on arbitrary surfaces without compromis- increasing interest in using it as the electrode for RRAM
ing its high conductivity. They fabricated their RRAM devices, several groups published their work to explore
device on various substrates including PET, a vial, further insights into this material. Qian et al. [65] fab-
medical wristband, glove, silicon wafer, coin and ricated an ultralow power RRAM device on a flexible
organic diode with one-step transfer process. Their PET substrate with integrated large area graphene elec-
study proves to be an important step forward for the trodes grown by CVD. The active layer of this memory
various possible applications such as epidermal elec- device was based on TiOx. These ultralow power devices
tronics and wearable computers. The obtained electri- did not lose their distinct features of large memory
cal and mechanical results were impressive, with window, high electrical endurance and retention when
switching ratio, retention time, and electrical endur- compared to TiOx RRAM devices with Pt electrodes.
ance of 106, 104 s and 107 respectively. Photographs of This group showed that the heterogeneous interface
as-prepared GMML under repetitive bending are between TiOx and graphene can both be used for reli-
shown in Figure 4(a). able nonvolatile memory purposes with nonlinear
Sci. Technol. Adv. Mater. 21 (2020) 152 M.M. REHMAN et al.

Figure 4. (a) Schematic diagram for the fabrication process of the memory label. Photographs of as-prepared GMML under
repetitive bending. Measurement arrangement of the GMML along with the chemical structures of PMMA and P3BT [64].
(b) Schematic drawing of a GMD [65]. (c) Schematic diagram of graphene charge trap memory GCTM device structure [90].

electrical characteristics and ultralow power loss oxide-based low-power nonvolatile RRAM device
(3 orders of reduction as compared to standard metal/ with a multilayer graphene (MLG) electrode. The key
oxide/metal structure). Schematic drawing of a GMD features of this device were low threshold voltage
device is shown in Figure 4(b). (<1 V), low operational current (<100 µA), low
Different groups contributed in various aspects of power loss (<100 µW), fast switching speed (~60 ns),
RRAM devices by solving different problems related to and high switching ratio (>104). This device also did
RRAM devices based on 2D materials. Yang et al. [66] not require any forming process before carrying out its
proposed an oxide-based RRAM device with functio- electrical characterizations. Chakrabarti et al. [69]
nalized graphene as electrode in a compact manner. reported an important device structure of an oxide-
These devices exhibited that functionalized graphene based memory device with graphene acting as both the
electrodes illustrate an intrinsic threshold switching. top and bottom electrode to sandwich the tri-layer of
The major issue of moderating the sneak paths in TiOx/Al2O3/TiO2. This device exhibited forming-free
RRAM devices to enhance data storage density and and highly nonlinear switching effect with a high
scalability was part of this work. Jeon et al. [67] studied switching ratio of 104. Moreover, the value of its cur-
the effect of using single-layer graphene (SLG) electro- rent compliance was extremely low (180 nA) hence,
des on the resistive switching properties of Pt/Al2O3 resulting in ultralow power consumption.
/TiO2/Pt memory device. They solved the problem of Tian et al. [70] exploited the fact that bandgap of
observing oxygen ion drift by using transparent gra- a bilayer graphene can be induced through applying
phene as the top electrode. They successfully proved vertical electric field. They reported the first ever gra-
that asymmetric resistive switching in this oxide-based phene electrode that was controlled by a gate terminal
memory device was due to the redox reaction by the with a tunable threshold voltage window. This gate
oxygen drift at the interface. This device with gra- terminal was used to induce bandgap in graphene
phene as the top electrode was electroforming-free that in turn allowed the electric field to penetrate the
with a low threshold voltage of ~1.5 V. Instead of active material of RRAM device hence, controlling the
using a SLG electrode, Zhao et al. [68] proposed an quantity of oxygen ions resulting in tuning resistive
Sci. Technol. Adv. Mater. 21 (2020) 153 M.M. REHMAN et al.

switching characteristics. Their research group to study the effect of replacing Pt electrode with
showed that by tuning the operating gate voltage, a graphene electrode without varying the functional
switching ratio of this graphene-electrode-based layer and top electrode. They verified that the power
RRAM device can be enhanced from 103 to 105. Lee losses and threshold voltages of SET-RESET condi-
et al. [71] explored the memory effect by exploiting the tions can be reasonably reduced by using a graphene
atomically thin (3 Å) characteristic of graphene by electrode. Ott et al. [76] also used SLG as the top
using it as the electrode of a vertical 3D structure electrode and achieved a switching ratio ten times
RRAM device. They claimed the nonvolatile RRAM higher than the metal electrodes with a low power
device with the lowest power loss and attributed it to density. This device displayed multi-level resistive
the ultrathin graphene electrode. A schematic cross- states with the possibility of storing more than one
section of the proposed RRAM and its I–V curve are bit in a single memory cell. The summary of all the
shown in Figure 3(c). Jang et al. [72] also reported above-discussed RRAM devices with graphene elec-
a highly uniform and ultra-low power polymer mem- trode along with their typical characteristics are listed
ory device with MLG like Zhao et al. who used MLG in Table 1.
electrodes for oxide-based memory device. This device
was based on poly(1,3,5-trimethyl-1,3,5-trivinyl cyclo- 2.1.2. Graphene as an active layer in RRAM devices
trisiloxane) (pV3D3) polymer and its cross-sectional RRAM devices with graphene as their active layer are
HR TEM image is shown in Figure 4(c). Recently, Jang the potential candidate for high-performance electro-
et al. [73] have used as MLG as the bottom electrode nics devices owing to their unique mechanical, optical,
for organic-inorganic perovskite on a flexible substrate electrical, and thermal properties because many
to resolve the problem of non-flexibility of perovskite advanced electronic devices need materials with atom-
materials. Their device displayed asymmetric and ically thin films and wide bandgap. Owing to the
reversible bipolar resistive switching behavior in the compatibility of graphene with the existing planar
operating voltage range of 0.68 V to −0.5 V. The stand- devices, it can be integrated with CMOS technology.
out feature of their device was its mechanical bend- Additional layers of graphene can act as the charge
ability as it was successfully tested for 1000 cycles at storage medium hence, resulting in higher retention
a bending radius of 4 mm without any noticeable time. Graphene can also offer high transparency, light
change in resistive states. The I–V characteristic curves weight, flexibility and low sheet resistance.
exhibiting its high electrical endurance are illustrated Doh et al. [81] and Shin et al. [82] are considered as
in Figure 10(b) with the inset showing the optical the pioneers of using graphene as the active layer of
image of the fabricated device. Unlike Jang et al., nonvolatile memory device. Both groups fabricated
who used MLG as the bottom electrode, Köymen a three terminal device but predicted the possibility
et al. [74] have reported SLG as the top electrode in of using graphene as the functional layer of a two
the memristive device with a configuration of terminal memory device. The graphene-based RRAM
platinum/TiO2/TiOx/graphene. The optoelectronic device of Shin et al. exhibited an ambipolar bistable
properties of graphene made it a viable choice as switching behavior with electrical endurance up to 100
a top electrode as it allowed to perform the in situ voltage sweeps and a small switching ratio of 13. The
analysis of the functional layer to a depth of 10 nm. It low value of switching ratio was attributed to the high
was due to this reason that they were successful in metal like conductivity of graphene. They also sug-
observing the changes in binding energies at the gested that active layer of graphene can be used for
atomic level to examine the drift of oxygen vacancies. storing multi-bit data and its switching ratio can also
Dugu et al. [75] fabricated two different RRAM devices be enhanced by inducing a bandgap in graphene

Table 1. Summary of graphene-electrode-based RRAM devices.

Bottom electrode Top electrode Active layer Substrate Switching ratio Endurance Retention (s) References
Graphene Graphene SWCNT PET 103 102 103 [60]
Al MLG PI:PCBM PET 106 104 104 [61]
Graphene Pt SiOx Si 102 80 104 [62]
ITO Graphene SiOx Glass 105 102 105 [77]
Graphene Graphene SiOx Plastic 106 102 – [77]
Graphene Al PMMA:P3BT PET 105 107 104 [64]
Pt SLG Al2O3/TiO2 Si 102 103 – [67]
Graphene Ti/Pt TiO2 PEN 102 102 106 [65]
Graphene Graphene TiOx/Al2O3/TiO2 – 104 102 104 [78]
MLG MLG Ta2O5–x /TaOy Glass 102 – – [66]
Graphene Graphene ZnO Si 103 50 – [79]
ITO Graphene Dy2O3 Glass 106 102 104 [68]
Graphene LSMO BTO Si/SiO2 103 103 – [80]
Al/MLG Cu pV3D3 Si/SiO2 20 102 – [72]
BLG Al AlOx SiO2 103 – – [70]
Sci. Technol. Adv. Mater. 21 (2020) 154 M.M. REHMAN et al.

layers. Although graphene-based switching device had using two similar polymers, they used two different
a drawback of lower switching ratio, owing to its high polymers of polystyrene (PS) and polyvinyl-carbazole
conductivity, it exhibited fast switching speed. Later, (PVK). This device was write-once-read-many-times
Doh et al. explored the effect of varying active layer memory (WORM) with a volatile behavior. The oper-
thickness on the resistive switching characteristics of ating voltage of this device was relatively high, in the
graphene-based three terminal devices. The major dis- range of ± 6 V, as compared to its predecessors; how-
advantage of this device was that its operational vol- ever, the achieved switching ratio of this hybrid device
tage was extremely high (>30 V) hence, resulting in was like the one reported by Song et al. (1 × 107). The
a huge power loss. They suggested that the value of fabricated device exhibited an average reproducibility
operational voltage can be decreased by reducing the of 64% as only 16 out of 25 cells operated appropri-
film thickness of graphene. Furthermore, the switch- ately. Schematic diagram of this hybrid bistable mem-
ing ratio of this device was also low, but its retention ory device is shown in Figure 5(a).
time was encouraging (2 × 103 s). Later that year, Singh et al. [86] proposed a bipolar
After the introduction of graphene as the active memory device based on the hybrid interface of copper
layer of a three terminal RRAM device with average oxide and MLG. Graphene acted as the blocking layer
properties, Son et al. [83] proposed a flexible bistable and a storage medium for oxygen anions. This device
memory device which was based on a three-layer required electroforming at an applied voltage of −0.84 V
structure of organic polymer embedded with gra- before its proper functioning. A high switching ratio of
phene. Graphene layers were sandwiched between 1 × 103 with a low operating voltage of ± 0.8 V was
the two layers of PMMA to provide current stability recorded for this hybrid device. No prominent decay in
and enhancing charge storage. This work can be con- its electrical performance was observed for 100 voltage
sidered as another breakthrough in the graphene- sweeps. It was deduced that the bilayer structure of Ti-
based RRAM devices as it exhibited extraordinary CuO-MLG-Cu exhibited bipolar resistive switching
characteristics of a nonvolatile memory device. The behavior while I–V curves of Ti-CuO-Cu displayed
switching ratio, electrical endurance and retention only a rectifying behavior. Their results clearly showed
time for this device were 1 × 107, 1.5 × 105 cycles that resistive switching in this hybrid structure was due
and 1 × 105 s, respectively. Based on the trend of to the inclusion of MLG without which their devices
obtained results, Son et al. predicted that their device displayed rectifying behavior instead of memory beha-
could have the lifetime of 10 years. The operating vior. The major drawback of this device was that it
voltage (± 2 V) of this device was also much lower as required a preliminary electroforming step that was
compared to the three terminal device based on gra- attributed to migration of oxygen anions from CuO
phene proposed by Doh et al. thus, resulting in less and induced electric field generation. Another impor-
power losses. The conduction mechanism of this tant result studied by this group was the impact of
device was attributed to the classic theory based on ambient and vacuum conditions on the resistive switch-
the formation of conductive filaments. Other than the ing characteristics. It was deduced that the values of
above mentioned fascinating traits of this hybrid reset voltage and current were significantly reduced to
device, it was the first memory device based on gra- 0.07 V and 1 mA as the operating conditions switch
phene active layer that was fabricated on a flexible from ambient to vacuum environment due to the avail-
substrate. These devices were bent up to a bending ability of more oxygen anions in vacuum conditions. I–
radius of 10 mm without any prominent decay in their V characteristics of this device during initial electro-
performance. This group suggested that embedding forming process displayed a forming voltage (VF) of
a complete layer of graphene in between the organic 0.84 V with current compliance current value of 1 mA
polymers instead of nanoparticles has added advan- as shown in Figure 5(b) with its schematic diagram
tages of reproducibility and higher flexibility due to shown in the inset.
uniform distribution of graphene. More articles related to the role of graphene as the
More research work was carried out on the gra- active layer in resistive switching devices were pub-
phene-based resistive switching device by Akdim et al. lished in the year 2012 owing to the promising char-
[84] who proposed the switching behavior of graphene acteristics of this 2D material as reported in the
nanoribbons. Although the switching characteristics previous years. Wu et al. [87] reported commendable
were not that impressive as compared to the work of results with a switching ratio of a single graphene sheet
Son et al., their distinctive contribution was to explore reaching up to 106. They listed an important drawback
the effect of uniaxial strain on graphene-based RRAM of using the hybrid active layer approach earlier
device. Wu et al. [85] proposed another hybrid mem- reported by Son et al. that due to the inclusion of
ory device similar in structure to the one proposed organic polymers, degradation time of these devices
earlier by Son et al. with a layer of graphene sheets can be quick due to the vulnerability of organic mate-
embedded in between two layers of polymers. The rials towards environmental conditions like humidity.
main difference in their structure was that instead of The main difference in this device from the previous
Sci. Technol. Adv. Mater. 21 (2020) 155 M.M. REHMAN et al.

Figure 5. (a) Schematic diagram of the hybrid bistable memory devices fabricated utilizing graphene sheets sandwiched between
polymer layers [85]. (b) Current-voltage characteristics of Ti-CuO-MLGCu sample during initial electroforming step showing
transition to low resistance state (LRS) at forming voltage (VF) of 0.84 V with a current compliance limit (ICC) of 1 mA. Inset
shows a schematic view of Ti-CuOMLG-Cu structure [86]. (c) Schematic diagram of the fabrication process for the two-terminal
devices. Uniform and large-area graphene film on SiO2 (300 nm)/Si substrates was used as starting materials; PMMA resist mask
was patterned by e-beam lithography; removal of the exposed graphene by reactive ion etching; removal of the PMMA overlay in
hot acetone; contact electrodes formation by a second-step e-beam lithography and Ti/Au (2 nm/30 nm) deposition by e-beam
evaporation; lifting-off of the metals for the two-terminal resistive switching devices [89]. (d) Geometry of the organic ferroelectric-
graphene memory device, where W is the width and L is the length of the graphene channel [96].

graphene-based memories was its structure as it was power losses. Like their previously reported sand-
a two terminal device with a planar structure instead wiched structure hybrid memory device based on PS/
of the typical sandwich structure. Due to a larger pla- graphene/PVK, this planar structured device also dis-
nar gap between anode and cathode (4 µm) as com- played WORM behavior. A second substrate at the top
pared to the sandwiched devices, its operating voltage of the active layer of graphene sheet was used to keep it
was on a higher side (±7 V) with a drawback of higher in contact with the planar electrodes during high
Sci. Technol. Adv. Mater. 21 (2020) 156 M.M. REHMAN et al.

current state. After the reports of Son et al. and Wu Kim et al. [90] recognized the supreme flexibility
et al. on the fabrication of memory devices based on present in the monolayer of graphene and fabricated
hybrid active layer, Ji et al. [88] developed another a memory device on a transparent polyethylene naph-
nonvolatile memory device based on the hybrid struc- talate (PEN) substrate. Only 8% loss in transparency of
ture in which layers of graphene were sandwiched PEN substrate was observed due to the presence of
between two layers of organic polyimide (PI) polymers highly transparent graphene layer on its surface. Their
with a configuration of PI/MLG/PI. The reason for device had a memory window of ~8.6 V with 30%
repeating this approach of embedding graphene layer retention/10 years. Their device displayed highly
in between two polymer layers was to produce high- stable and repeatable bistable resistive states under
quality results. The switching ratio achieved by this successive compressive and tensile stress cycles. The
group was a little lower compared to some earlier proposed RRAM device by Kim et al. based on gra-
reports, but it was still superior to memory devices phene had a potential to be used as a flexible and
based pure graphene (>106) with a decent retention transparent electronic memory device with high
time (104 s). Although all these reported results with endurance under flex.
hybrid active layer approach achieved results that were Lai et al. [91] proposed a memory device based on
close to each other, the major difference of this device graphene nanoflakes (GNF) assisted macromolecular
as compared to the previously reported devices was composite memory device with a low operating vol-
the selection of a different set of organic polymers and tage. They proposed a solution-processable functional
electrodes. The memory device by Ji et al. exhibited material that was a blend of GNF and poly(vinyl
excellent cell to cell distribution that was verified by alcohol) (PVA). This work was the first report on the
their cumulative probability plots. The fabrication of direct integration of graphene into any RRAM device
this device was relatively more difficult and complex as through a low-temperature and single-step processing
it employed a greater number of steps and was time- technique. Operating voltage of this device was rela-
consuming. They also proved that memory behavior tively low (−1.3 V to +1.4 V) with a remarkable elec-
in this hybrid device was due to the charge trapping trical endurance of 107 voltage cycles. The major
ability of MLG sandwiched layer. A schematic dia- advantage of the graphene-based RRAM device pro-
gram of the fabricated memory device is shown in posed by this group was to achieve excellent switching
Figure 5(d). properties through an extremely simple and cost-
He et al. [89] proposed another remarkable gra- effective solution-process method with an active area
phene-based memory device with extremely promis- of 3 mm2.
ing characteristics of a nonvolatile RRAM. Their Another important work was reported by
study was comprehensive and instead of using the Shindome et al. [92] as they thoroughly investigated
typical sandwiched structure, He et al. also choose the conduction mechanism taking place in graphene-
the planar structure for their device just like the two based RRAM devices by varying the number of gra-
terminal memory devices of Wu et al. A lot of new phene layers, changing device size and varying the
insights were created in this work such as instead of contact metals. Monolayer, six-layer and eight-layer
only two resistive states, this RRAM device had mul- graphene was used as the functional layer in the form
tilevel resistive switching with a total of five different of a planar structure. They used standard microscopy
states. The switching speed and yield of this device techniques of atomic force microscope (AFM), elec-
were the other two prominent features with the tron energy-loss spectroscopy (EELS) and cross-
values of 500 ns and 98%, respectively. Moreover, sectional TEM to examine the conduction mechanism
the gap between anode and cathode was in nanoscale of their memory device and concluded that it is based
instead of microscale hence resulting in extremely on the chemical bonding-state change of graphene
low operating voltage and reduction of power losses nanoribbons. Yang et al. utilized the highly transpar-
that was a drawback of the microscale planar mem- ent property of graphene to fabricate a fully transpar-
ory device proposed earlier by Wu et al. The draw- ent RRAM device based on ZNO active layer. They
back of this RRAM device was that it required an proved that combining graphene with ZNO had
electroforming step (Vbreak = 5.5 V) before applying a significant impact on the resistive switching proper-
a double voltage sweep across its electrodes to cause ties with not only enhanced transparency but also
electrical breakdown. Unlike the typical RRAM reversible resistive switching by suppressing the sur-
memory devices, this device neither showed unipolar face effects. Graphene absorbed only ~2% of incident
or bipolar memory behavior; instead, it exhibited light.
nonpolar resistive switching behavior independent Berzina et al. [93] reported another RRAM device
of voltage polarity. A schematic diagram of the fab- based on the thin film of polyaniline-graphene compo-
rication process for the two-terminal device is shown site by applying the Langmuir-Schaefer method. They
in Figure 5(c). analyzed the variations in electrical properties such as
Sci. Technol. Adv. Mater. 21 (2020) 157 M.M. REHMAN et al.

S-like I–V curves and increased operating voltage range organic memristor with its I–V characteristics are
that were associated with accumulation of charges in shown in Figure 6(b). The trend of embedding metal
graphene sheets. A schematic representation of the oxide nanoparticles in a polymer layer on a SLG was

Figure 6. (a) Schematic 3-dimensional structure of memory device with SnO2 nanoparticles embedded in BPDA-PDA polyimide
layer on SLG contact layer. I–V curves of the resistive switching memory with SnO2 nanoparticles embedded in BPDA-PDA
polyimide layer on SLG contact layer [94]. (b) Circuit and schematic diagram of the GQD RRAM device along with the simulated I–V
characteristics of the device [113]. (c) Electric hysteresis loop of graphene ferroelectric memory. R as a function of top gate voltage
for the graphene-ferroelectric sample and Schematic of graphene floating gate FLASH memory [97]. (d) The magnifying schematic
structure of the ITO/PS/SLG/PMMA/SLG/PMMA/Al memory unit cell. I–V curves of ITO/PS/SLG/PMMA/SLG/PMMA/Al device and I–V
curve of ITO/PMMA/SLG/PMMA/SLG/PMMA/Al device [101].
Sci. Technol. Adv. Mater. 21 (2020) 158 M.M. REHMAN et al.

pioneered by Lee et al. [94] when they reported an limited current mechanisms were responsible for resis-
interesting active layer of SnO2 embedded BPDA- tive switching effect in these devices. These graphene-
PDA polyimide on SLG. However, this experiment based tri-layer memory devices showed highly stable
was not successful as it displayed a maximum switching resistive switching effect in four different states with
ratio of only 7.9 with a relatively higher operating a high retention time (104 s) and good electrical endur-
voltage of ± 5 V. A schematic diagram and I–V curve ance (150 biasing cycles). Although such a device struc-
of the proposed RRAM device are shown in Figure 6(a). ture with a sandwiched graphene layer between two
A significant result was reported by He et al. [95] as they polymers layer had already been used by Son et al.,
explored the tunable electroluminescence (EL) effect in Wu et al., and Ji et al., but the major differences of the
the planar structure RRAM device based on proposed RRAM device by Khurana et al. is the differ-
graphene/SiO2. They concluded that this relation ence in selected polymer material causing a multilevel
between the EL and resistive states can be used to resistive switching with a distinguished advantage of
distinguish between both. Their device could be used high-density data storage.
for the possible applications of optical communications. Wu et al. [101] proposed a multilevel resistive
Wang et al. [97] published a research news on the switching behavior in a different device structure
progress of graphene-based nonvolatile memory devices based on single-layer graphene. The proposed device
in which they acknowledged the substantial impact of configuration consisted of five layers with the config-
graphene on resistive switching memory devices but uration of polymer/SLG/polymer/SLG/polymer sand-
emphasized on further development in this field to wiched between two metal electrodes. Although the
reach a sizeable impact. They suggested that in graphene- switching ratio (104) of this device was more than the
based memory devices one should be careful about one reported by Khurana et al., it had a drawback of
response time and power consumption of their memory complex structure and use of excessive materials as it
devices. Furthermore, they pursued that low cost and contained two extra layers. A schematic diagram and
highly reliable fabrication methods are the need of hour obtained I–V curve of the fabricated device with
for achieving high throughput memory devices. multi-resistive states are illustrated in Figure 6(d).
Furthermore, it was suggested by Wang et al. that a lot Furthermore, it consisted one less resistive state
of research efforts were required to explore the nanoscale hence resulting in lower data storage. Atab et al.
memory devices to realize their integration with CMOS [102] illustrated that the memory behavior can be
systems. They rightly raised the hope of a flexible mem- enhanced in ZnO-based RRAM device by embedding
ory device based on the promising results achieved by the graphene nanoplatelets in it. Graphene nanoplate-
graphene RRAM cells. Electric hysteresis loop of gra- lets acted as the charge trapping material that also
phene ferroelectric memory and its schematic diagram helped in lowering the threshold voltage hence result-
are shown in Figure 6(c). Rocci et al. [98] reported a nano ing in low power consumption due to its high con-
hybrid based memory device based on a few layer gra- ductivity value. Inclusion of graphene nanoplatelets
phene deposited on the planar structure of manganite also enhanced the retention time up to 10 years with
electrodes grown on SrTiO3 substrate. These devices 25% data loss. Later that year, Cristea et al. [103]
exhibited hysteric transport and rectifying resistive reported the impact of embedding graphene quantum
switching behavior as depicted from their I–V curves dots (GQDs) in the organic polymer of poly(3-hex-
with the ability of working properly in the temperature ylthiophene) (P3HT) on its resistive switching char-
range of 10 K to 300 K. The memory effect in the single acteristics. They also verified that GQDs act as the
crystalline multilayer and monolayer-graphene-based charge trapping material in the polymer-based mem-
nano switches was explored by Li et al. [99] These devices ory device. Inclusion of GQDs greatly enhances the
had the ability to be operated at a pull-in voltage of 1 V hole transport. The P3HT-GQDs-based RRAM device
with a high switching speed of ~100 ns and an electrical had a planar structure with a long conduction channel
endurance of 5000 voltage sweeps that was better than of 8 µm and two Au electrodes acting as anode and
other reported graphene-based nanoelectromechanical cathode.
resistive switches. Monolayer graphene-based device Our research group explored the resistive switching
exhibited lower pull-in voltage while the bilayer- effect in the nanocomposite of poly(4-vinylyphenol)
graphene-based device showed higher speed. (PVP) and graphene nanocomposite thin films. The
Impressive switching results in a tri-layer structure obtained results were significant in a way that different
of graphene flakes sandwiched between two layers of approaches of device structure were studied including
polyvinylidene fluoride (PVDF) polymer were reported blended and non-blended functional thin film. It was
by Khurana et al. [100]. This tri-layer geometry was concluded that blended functional thin film of PVP-
further sandwiched between indium tin oxide (ITO) graphene exhibited better stability and higher switching
bottom electrode and Pt top electrode to exhibit multi- ratio as compared to using these two materials as
level resistive switching for storing high data density. a bilayer [104]. As a continuation of exploring the resis-
They found that Fowler-Nordheim and space charge tive switching effect of this nanocomposite, it was
Sci. Technol. Adv. Mater. 21 (2020) 159 M.M. REHMAN et al.

fabricated on a flexible PET substrate to evaluate its graphene layer was that it reduced the random rupture of
mechanical robustness. The devices were bendable up filaments hence resulting in stability of the obtained
to a bending diameter of 1.8 cm for 50 bending cycles results. Furthermore, the value of threshold voltages
without any prominent decay in its performance. Lee was also reduced that resulted in lower power losses.
et al. [105] illustrated that the resistive switching char- The whole fabrication process and the obtained electrical
acteristics of an oxide layer can be enhanced by simply characteristics of the proposed RRAM device are
inserting a defective graphene layer. The impact of this illustrated in Figure 7(b). The memory effect in the

Figure 7. (a) I–V characteristic of memory-only (1 R) device and its Log–log plot of I–V characteristics at 300 K [107]. (b) Resistive
switching characteristics of MGIM structures compared with a conventional MIM structure. Illustration of fabrication process for
MGIM structure. D-graphene is made before it is transferred to device. Initial current-voltage characteristics of the MGIM and
conventional MIM structures. Cumulative probability of switching voltages, Vset and Vreset, for MGIM structures with d-graphenes
irradiated with Ar+ ions at kinetic energies of 240 eV (MGIM240), 250 eV (MGIM250), 260 eV (MGIM260) and 270 eV (MGIM270) as
well as a MIM structure [105]. (c) Schematic of the graphene-inserted memristor structure where oxygen ions transport only
through a nanopore created in the graphene layer, forming a CF with oxygen vacancies with I–V curve [109].
Sci. Technol. Adv. Mater. 21 (2020) 160 M.M. REHMAN et al.

polymer-suspended graphene nanoplatelets was reported use graphene as either an active layer or electrode in
by Kang et al. [106]. Although the device exhibited reli- the RRAM device.
able performance, its switching ratio was only 10. Pan et al. [113] fabricated a two-terminal and
Ji et al. [107] investigated the resistive switching a three-terminal RRAM device based on GQDs with
behavior of organic polymer after mixing it with the resonant tunneling and volatile memory behavior.
nanocomposite of ZnO-GQDs and exhibited their use Addition of the third terminal was useful to modify
as an active element in combination with a diode-one the resistive switching states. They also used branch
resistor. They obtained average memory results with and load resistors with this device to obtain different
a high reliability by using ZnO-GQDs nanocomposite hysteresis characteristics. Introduction of third gate
acting as charge trap sites. The architecture used in terminal induced more logic states for each memory
this work can be very effective to reduce or eliminate cell. Kapitanova et al. [114] have explored the photo-
the cross talk between consecutive memory cells. They memristive effect in the heterojunction of self-
tested their devices through encoding a complete word assembled nanoscale graphene oxide/graphene. These
by using the standard ASCII code. A schematic dia- devices were photosensitive with ultra-high capacity
gram, its I–V curve and double logarithmic I–V curve nonvolatile memory and multilevel resistive switching.
are illustrated in Figure 7(a). These devices were functional both in dark conditions
Kurkina et al. [108] explored the switching effect in and under the effect of light. Recently, Shehata et al.
the partially fluorinated graphene thin films. Lee et al. [115] used a nanocomposite of graphene and TiO2 in
[109] found the ionic transport mechanism in memory a planar device as the functional layer of RRAM device.
devices of graphene sheets with engineered nanopores. Nanoparticles of TiO2 were uniformly dispersed on the
They tuned the resistive switching characteristics of surface of graphene sheets by using UV irradiation that
tantalum oxide material by inserting a graphene film helped in enhancing the trap sites for the electrons. The
with engineered nanopores. The role of graphene was I–V characteristic curve and its schematic diagram of
to block redox reactions and ionic transport. The effect the planar structure are shown in Figure 10(a).
of different sized nanopores was also studied and it was Liu et al. [116] fabricated a graphene-based RRAM
depicted that value of LRS and HRS were directly device with extremely low power consumption of less
proportional to the nanopore size of graphene. The than a femtojoule by using the weak van der Waals
effects of annealing on the ion migration due to engi- forces of its surface and suggested their device for the
neered nanopores of graphene were also studied at possible application of neuromorphic computing sys-
a high temperature of 300°C. Schematic of the gra- tems. Sun et al. [117] successfully eradicated the nega-
phene-inserted RRAM structure in the graphene layer, tive set response of ITO/methacrylate epoxy resin
along with its I–V curve is illustrated in Figure 7(c). (MER)/Al-based RRAM device through the addition
Mannequin et al. [110] reported the resistive switching of graphene blocking layer. This result was achieved
effect in the interface of graphene and HfO2. They due to the resistant nature of graphene layer that does
compared the device performance of HfO2 device not allow the conductive filament to diffuse into the
with and without graphene layer between the top inert ITO electrode. Wu et al. [118] realized that by the
electrode and found out that an immense improve- addition of graphene layer in a binary transition
ment in the retention of ON state occurred by adding metal-oxide-based memory device, its volatile beha-
graphene layer due to its oxygen reservoir effect. vior can be changed to nonvolatile with high stability
A very different and significant study was reported and the ability to store multiple bits in a single cell.
by Ueda et al. [111] in which they reported the photo- Yalagala et al. [44] developed a memory device based
controlled memory effect in the heterojunction of on the composite of MgO-PVP-Graphene and used it
graphene/diamond with multilevel resistive states. for the security-based applications. The distinctive
The obtained results showed that only blue and violet feature of this study was that the whole device was
light had the ability to switch the resistive states with water dissolvable in less than 3 s owing to the high
a maximum switching ratio of 103 even at a high hydro-dissolution rates of PVP and MgO. They pro-
temperature of 200°C. These results clearly illustrated posed a novel feature of destroying the memory from
that the heterostructure of graphene/diamond have their device wirelessly from a remote place. The sum-
the potential to be used as either a photo switch or mary of all the above discussed graphene-based
for photometry purposes. Among those regular RRAM devices along with their typical characteristics
research papers, a review article was written by Rani are listed in Table 2.
et al. [112] to discuss the conduction mechanism tak-
ing place in the nonvolatile RRAM devices based on
2.2. Role of reduced graphene oxide/graphene
graphene due to the fact that researchers were taking
oxide in RRAM devices
keen interest in using it as the active material for
RRAM devices. This article contains ample amount Nonvolatile and bistable memory devices based on
of information for those readers who are interested to Graphene Oxide (GO) have prompted great interest
Sci. Technol. Adv. Mater. 21 (2020) 161 M.M. REHMAN et al.

Table 2. Summary of RRAM devices with graphene as the active layer.

Bottom electrode Top electrode Active layer Substrate Switching ratio Endurance Retention (s) References
Ti/Au Au Few-layer graphene (FLG) Si/SiO2 – – 103 [81]
Cr/Au Al Graphene SiO2 – 102 – [82]
ITO Al PMMA/graphene/PMMA PET 106 105 10 years [83]
Al ITO PS/graphene/PS Glass 107 3 106 [85]
Al ITO PVK/graphene/PVK Glass 104 5 102 [85]
Cu Ti MLG/CuO Si 103 102 – [86]
Ti/Au Ti/Au Graphene SiO2 103 104 105 [89]
Al Al PI/MLG/PI Si 106 – 104 [88]
ITO ITO Graphene Glass 106 – 104 [87]
Ti/Cr/Au Ti/Cr/Au Graphene nanoribbons Si/SiO2 106 102 103 [92]
ITO Ag GNF:PVA Glass 102 107 104 [91]
SiO2 Al Graphene/BPDA:PDA Si 7.9 – 104 [94]
Ti/Au Ti/Au Graphene Si/SiO2 105 – – [95]
ITO Pt PVDF/Graphene/PVDF Glass 105 50 104 [100]
Au Au P3HT:GQDs Si – – – [103]
ITO Al PS/SLG/PMMA/SLG/PMMA Glass 104 – 104 [101]
ITO Ag PVP-Graphene Glass 5 20 – [104]
Al Al ZnO-GQDs SiO2/Si 103 102 104 [119]
ITO Ag Graphene/PVP PET 35 40 102 [120]
NiO Pt Graphene/NiO Ni/Cu 102 102 105 [105]
Pt/Ti Al ZnO-GQDs SiO2 103 103 104 [45]
ITO Au Al2O3/Graphene Glass 103 102 104 [75]
ITO Al Graphene/MER PET 103 40 104 [117]
Pt Ag Graphene/HfOx – 107 103 103 [118]
Ag ITO MgO-PVP-Graphene PET 103 102 104 [44]

due to its high optical transparency, high flexibility, explored the nonvolatile memory behavior and high
low cost, easy fabrication, environmentally friendly flexibility of RGO RRAM device. This was the first
nature, controllable physical and chemical properties, ever RRAM based on RGO that was fully solution-
for future transparent and flexible electronic devices. processed and flexible. In this RRAM device apart
Vasu et al. [121] in 2011 reported the unipolar resistive from the active layer, both electrodes were also made
switching effect in the ultra-thin films of reduced of RGO making the device a low cost and environ-
graphene oxide (RGO) on a rigid glass substrate with ment-friendly RRAM. Like earlier reported RRAM
a high yield of >99%. The obtained switching ratio of devices based on RGO, this RRAM device also had
105 with extremely fast switching speed of 10 µs was an undesired feature of high threshold voltage
much higher than most of the graphene-based (GO) (±14 V). The flexibility of the RRAM device was tested
memory devices. This RGO-based RRAM device was for 1000 bending cycles without any prominent decay
compared with a MWCNT-based RRAM device in its performance. Seo et al. reported the rewritable
whose switching ratio was much lower (400) with and nonvolatile memory behavior in the partially
huge power losses as its threshold voltage was reduced graphene oxide doped with nitrogen. They
20 V. Later, Cui et al. [122] studied the nonvolatile mainly targeted the drawback of RGO-based memory
memory behavior of the covalently bonded gold nano- devices, i.e. uncontrollable oxygen functional groups.
particles embedded in RGO thin film. They also In order to induce stability in the obtained results, the
observed the effect of device structure on its properties functional layer was doped with nitrogen. They
by fabricating both sandwiched and planar structure deduced that the memory behavior of partially
device with the same active layer. The devices dis- reduced GO was highly dependent on the concentra-
played stable results of multiple write-read-erase tion of doped nitrogen and the RRAM device comple-
cycles (103 cycles) with high retention time (700 s) tely lost its memory behavior with decreased
while the switching ratio exhibited by the sandwiched concentration of nitrogen atom.
structure device was higher than the planar structure. Han et al. [125] investigated the tunable memory
A schematic diagram and I–V characteristic curve of characteristics by engineering the energy band
the proposed device is shown in Figure 8(c). through controlled doping of RGO. As the functional
Hu et al. [123] in 2012 published a research article layer was doped with gold chloride (AuCl3), a gradient
in which they controlled electron transfer effect in increase in the value of RGO work function was
RGO that was noncovalent functionalized with thio- noticed. They demonstrated a flexible RRAM device
nine. The obtained results were promising with a fast with a large memory window and controllable thresh-
switching of <5 ns, switching ratio of 104, and reten- old voltages. The basic fabrication process is shown in
tion time of 105 s. This study provided a new method Figure 9(a), along with a schematic diagram of device
for regulating the electronic characteristics of gra- structure, and its optical image is shown in Figure 9(b).
phene for RRAM devices. In the same year, Liu et al. Ho et al. [126] studied the switching mechanism of
[124] published an important article in which they RGO active layer through impedance spectroscopy.
Sci. Technol. Adv. Mater. 21 (2020) 162 M.M. REHMAN et al.

Figure 8. (a) Schematic diagram depicting the basic fabrication process of reduced GO-based memory devices. Patterning APTES
SAM on the SiO2 substrate using photoresist patterns. Assembly of GO pieces onto APTES layer in solution. Removal of photoresist
and fabrication of electrode. Reduction of graphene oxide. Deposition of Al2O3 via ALD, patterning APTES and OTS SAM layers on
the Al2O3 layer using microfabrication, and the assembly of Au NPs onto the APTES SAM regions. Here, OTS was used for
passivation. Deposition of a thick Al2O3 layer followed by the fabrication of top-gate electrodes [128]. (b) J–V characteristics and
stability tested in either ON or OFF state under stimulus by read pulses of a 0.16-mm2 ITO/TPAPAM-GO/Al device. Inset: schematic
diagram of the single-layer memory devices [129]. (c) I_V characteristics on log scales of the vertical device (i.e. Al top-electrode
/AuNP-fRGOs hybrid material film/ITO bottom-electrode). Noticeable nonlinear hysteresis is observed in all devices [122].
(d) Current density/voltage (J/V) characteristics of a 0.16 mm2 Al/PFCF-RGO/ITO device; the symbols indicate sweeps 1–8 [130].
(e) Three-dimensional schematic illustration of RGO-based organic transistor memory device. Photograph of the fabricated flexible
RGO transistor memory devices. SEM image of RGO on the APTES-coated substrate [131].
Sci. Technol. Adv. Mater. 21 (2020) 163 M.M. REHMAN et al.

Figure 9. (a) Schematic diagram depicting the basic fabrication process of AuCl3 doped RGO floating gate memory device. (b)
Fabricated device structure for memory device with inserting AuCl3 doped RGO floating gate and molecular structure of pentacene
and F16CuPc. The optical image of the flexible memory [125]. (c) I–V characteristics of the ITO/RGO/ITO device; the inset shows the
ITO/RGO/ITO structures [127].

They compared the results of RGO with pristine GO However, despite better operating voltage and reten-
active layer to figure out the possible conduction tion time, its switching ratio was inferior to the pre-
mechanism responsible for resistive switching beha- viously reported RRAM devices. Later that year, Kim
vior. They deduced that memory behavior in RGO is et al. [127] published an article and reported the multi-
due to the oxidation and reduction reactions taking level resistive switching behavior in reduced graphene
place at the top interface between oxygen migration in oxide sandwiched between two ITO electrodes by fab-
the active layer and Al electrode. The retention time ricating a transparent memory cell. This RRAM device
for this device was >106 s that was the highest among exhibited an optical transmittance exceeding 80% with
all other RGO-based memory devices reported till then 00, 01, 10, and 11 resistive states. Multilevel resistive
and the operating voltage was also low (±4 V). states were achieved by varying the amplitude of
Sci. Technol. Adv. Mater. 21 (2020) 164 M.M. REHMAN et al.

Figure 10. (a) Current-voltage characteristics of graphene oxide (GO) and graphene oxide/TiO2 (GOT) devices; the inset is the
planar structure of prepared devices [115]. (b) Resistive switching characteristics after repeated bending tests up to 1000 cycles at
R = 4 mm [73]. (c) 3D 1S1 R scheme, cell structures and typical I–V characteristics of the control, CDG, DDG devices. Schematic
illustrations of the 3D 1S1 R crossbar array where the RS memory (resistor) and the accessory selector are indispensable
components in an operational cell. Cell structure and typical RS I–V characteristic comparison of the control (first line), CDG
(second line) and DDG device (third line) under various ICC. The maximum volatile ON-state currents are 100, 0.1 and 500 [145].

applied pulse voltage from 2 V to 7 V. The device the nonvolatile and bistable memory effect in the GO
displayed high endurance and retention time of 105 s layer functionalized by a conjugated polymer of tri-
each at a temperature of 85°C. The I–V characteristics phenylamine-based polyazomethine (TPAPAM) with
of this multiresistive RRAM device with its schematic the device configuration of ITO/TPAPAM-GO/Al.
diagram are shown in Figure 9(c). Myung et al. [128] The obtained I–V characteristics were commendable
proposed an ambipolar RRAM device with a planar with an electrical endurance of 108 and a switching
structure and RGO embedded with Au nanoparticles ratio of 103 as illustrated in Figure 8(b). Zhang et al.
as its active layer. The addition of Au nanoparticles [130] proposed another RGO-based RRAM device
caused a huge increase in the size of hysteresis win- grafted with conjugated polymer poly- [{4,4’-(9H-
dow. This device exhibited an n-type and p-type beha- fluorene-9,9-diyl)bis(N,N-diphenylbenzenamine)}
vior in the negative and positive applied bias region, {4-(9H-carbazol-9-yl)benzaldehyde}(9,9-dihexyl-9H-
respectively. The fabrication procedure of this device is fluorene)] (PFCF-CHO) with a device configuration of
illustrated in Figure 8(a). Zhuang et al. [129] reported ITO/PFCF-RGO/Al. The switching ratio (104) and
Sci. Technol. Adv. Mater. 21 (2020) 165 M.M. REHMAN et al.

electrical endurance (108) of this device were the both memory behaviors was attributed to the change
standout features of this device. The I–V curves of in the barrier at the interface of GO and POM. Their
first 8 voltage cycles along with a schematic diagram work was an important contribution as it discussed a lot
are shown in Figure 8(d). Rani et al. [131] proposed of insights about the tunable memory characteristics
a RGO-based memory device with a transistor-like and design of GO-based hybrid memory devices.
device structure on a flexible substrate. Their low Zhang et al. [135] also proposed a hybrid memory
cost and facile fabrication process-based memory device based on GO and a polymer with the configura-
device exhibited nonvolatile memory behavior with tion of Al/PVK-AZO-GO/ITO. The main contribution
an electrical endurance, and retention time of 102 of this team was also based on suggesting the new
and 105 s, respectively. The device structure in the design of a GO-based nonvolatile memory device with
form of its schematic diagram along with its optical the feature of multiple storage. Another hybrid compo-
image is illustrated in Figure 8(e). site based on the mixture of a polymer (PI) and GO
Lin et al. [132] investigated the resistive switching layer was proposed by Choi et al. [136] by using
behavior of ZnO RRAM device with a capping layer of a simple fabrication technique that illustrated
RGO layer. They depicted from the obtained results a WORM behavior. The switching ratio achieved by
that RGO insertion brought stability to the memory this group was extraordinary as it exceeded 108 which
behavior of ZnO with a better switching ratio of 105. was ascribed to the high charge storage capacity of GO.
They stated that the layer of RGO played the role of Wan et al. [137] also proposed a hybrid memory
oxygen reservoir in this RRAM device hence, making device of GO layer but instead of using a polymer as
the transportation of oxygen ions easy. Another the second material, they preferred to use a perovskite
important advantage of inserting RGO layer was that oxide of strontium titanate (SrTiO3). SrTiO3 nanopar-
it restricted the oxygen vacancies to react with Al ticles were embedded in the functional layer of RGO
electrode. This device displayed the highest electrical and the obtained results illustrated an interesting
endurance of 108 voltage sweeps as compared to all mechanism in which the resistive switching behavior
other RRAM devices based on reduced graphene of the fabricated device changed from digital to ana-
oxide. In the same year, Midya et al. reported the log. The digital memory effect was due to the forma-
resistive switching effect in a different composite of tion and rupturing of filaments inside the functional
reduced graphene oxide blended with PVA polymer layer of SrTiO3 on the application of applied voltage;
embedded with Au nanoparticles. Although they however, the memory effect was tuned to analog after
obtained average results, their work opened the gates the addition of RGO layer as Ag filaments were not
for other researchers to explore the memory behavior easily broken on the application of negative voltage.
in other hybrid nanocomposites of RGO. Kim et al. [138] proposed another important study for
Finally, Pradhan et al. [133] published an important understanding and improving the conduction
article in which they resolved the problem of high mechanism of in GO-based memory devices due to
power losses in RGO-based RRAM devices by fabri- their potential of future nonvolatile memory device.
cating a low power nonvolatile RRAM device with They used an in situ TEM technique to observe the
a threshold voltage even less than 1 V. Earlier the chemical and physical changes in the bonds of nanos-
lowest operating voltage for RGO-based RRAM device cale GO sheets and oxygen owing to the application of
was 4 V achieved by Ho et al. They also examined the biased voltage. They clearly observed that a conical-
effect of device size, film thickness and scan time on shaped dynamic conductive graphitic channel grows
the switching properties of the as fabricated memory from the upper region with more oxygen region to the
device. The resistance value had a direct relation with lower region with less oxygen. Qi et al. [139] also
temperature owing to which they deduced that the worked on the conduction mechanism of GO-based
conduction mechanism of this device was due to the memory device and engineered a new way to change
formation and rupture of metallic filament. the volatile memory nature of GO-based memory
Cao et al. [134] mixed the 2D material of RGO with device to a stable bipolar resistive switching by embed-
a novel and highly soluble polymer-based on organo- ding oxidized carbon quantum dots (OCQDs). They
phosphorus to form a device configuration of Al/ also attributed the switching mechanism in their
PTPEB-g-RGO/ITO. Although Midya et al. also device to the migration of oxygen vacancies.
reported the resistive switching effect in a hybrid They verified a definite increase in the retention of
nanocomposite of RGO with a polymer earlier on, low resistive state due to the addition of OCQDs after
Cao et al. achieved superior results with a switching studying the effect of temperature on GO-OCQDs
ratio exceeding 104 and a retention time of 104 s. nanocomposite-based functional layer. They enhanced
Chen et al. [42] reported a dual nature memory retention was due to the increase in barrier height that
device based on the bilayer structure of GO and poly- restricted the migration of oxygen. Such devices can be
oxometalates (POMs) with the features of bipolar resis- very useful in reducing the power consumption of
tive switching and WORM. The transition between future nonvolatile memory. Srivastava et al. [140] used
Sci. Technol. Adv. Mater. 21 (2020) 166 M.M. REHMAN et al.

the finite element modeling (FEM) to observe the effect content of GO, i.e. 5 wt% showed completely conduc-
of GO and RGO as a bilayer in TiO2 based memory tive behavior without any memory effect. Zhao et al.
device. Their study showed that GO acted as the chan- [145] solved the problem of current retention in car-
nel for the formation of filaments; however, RGO acted bon-cation-based memory devices that is associated
as an electrode instead of a functional layer. This work with the deteriorating stability of conductive fila-
contained experimental results backed by the theoreti- ments. The deterioration occurs by decreasing the
cal results obtained by the simulation. Saini et al. [141] compliance current value hence causing the high
studied the effect of annealing on the bipolar resistive amount of power losses. This group solved this pro-
switching behavior of GO layer at different values of blem by controlling distribution of centralizing and
temperature ranging from 100°C to 400°C. They also decentralizing the conductive filaments. The managed
used the TEM technology and observed the formation to control the size and quantity of these filaments by
of aluminum filament from top to bottom electrode. inducing a defective graphene layer in this structure as
The maximum switching ratio of 104 was observed at an shown in Figure 10(c). The summary of all the above-
annealing temperature of 200°C. discussed GO/RGO-based RRAM devices along with
Shi et al. [142] observed the complementary resistive their typical characteristics are listed in Table 3.
switching in GO-based memory device. They studied
the effect of varying compliance current on the quantity
3. Role of MoS2 in RRAM devices
of oxygen defects for the formation of conductive chan-
nels within the functional oxide layer. They reported The potential of NVM effect in MoS2 was first
that complementary resistive switching behavior was explored by Liu et al. [176] when they illustrated the
only observed when the magnitude of rest voltage was resistive switching effect in a hybrid nanocomposite of
greater than that of set voltage. Songkeaw et al. [143] MoS2-PVP sandwiched between RGO and aluminum.
fabricated a completely transparent WORM device by It was observed that the bipolar resistive switching
forming a composite of poly (ethylene-co-vinylacetate) shown by this device was due to the trapping and
with GO. The optical transmittance of the formed detrapping of charge carriers from MoS2 flakes
memory device in the visible light was 60% with blended in organic polymer (PVP). Although MoS2-
a high switching ratio of 105 and good retention time PVP memory device exhibited average performance
under variable temperature conditions. with a switching ratio of 102 and operating voltage of
Recently, Kim et al. [55] fabricated a multi-stacked ±5 V, it illustrated the potential of another 2D material
RRAM device based on the sandwiched composite to be used as the active material for flexible memory
layer of GO-PVA between the two insulating poly- device of the future.
meric layers of PVA. They also observed the effect of After the successful experiments by Liu et al., Xu
changing temperature like Saini et al. on their device et al. [177] reported another important study in which
by changing its value from 100°C to 200°C. instead of MoS2 flakes blended with another material,
Interestingly the switching ratio observed by this they used only MoS2 in the form of nanospheres as the
group at 200°C was also 104 as reported by Saini active layer sandwiched between RGO and ITO elec-
et al. earlier. This work seemed to be not contributing trodes on a rigid substrate of Si. Unlike the average
much in the advancement of achieved results based on switching properties shown by the RRAM device
GO RRAMs with a drawback that it used several layers based on MoS2 flakes, MoS2 nanospheres exhibited
hence, increasing the cost and difficulty level of device excellent memory behavior with a higher switching
fabrication. In the ongoing year, Li [144] has also ratio (104), lower operating voltage (±2 V) and a reten-
shared his tunable results of solution-processable GO- tion time of 104 s. These achieved results were attrib-
based memory device mixed with the polymeric layer uted to the unusual electrical properties of nanosphere
of polymethyl methacrylate (PMMA) by using the assemblies. Later, Sun et al. [178] explored the bipolar
theory of charge trapping. The tunable characteristics resistive switching effect in the single layer of MoS2
of this device were controlled by varying the concen- sandwiched between fluorine-doped tin oxide (FTO)
tration of GO (0.5–5 wt%) in the insulating layer of and Ag. Although the switching ratio of this device
PMMA. They observed that 0.5 wt% of GO did not was lower as compared to the one reported by Liu
have any impact on the OFF state characteristics of et al., the device exhibited an extremely low operating
PMMA-based RRAM device; however, the overall voltage (±0.4 V) with electrical endurance up to 100
conductivity of the active layer began to increase. voltage sweeps that is highly desirable as it reduces the
Interestingly the device with the higher content of overall power consumption. Bessonov et al. [179] pro-
GO (2 wt%) showed a lower switching ratio that can posed a solution-processed bilayer structure of MoS2
be attributed to the fact that by increasing the content and molybdenum oxide (MoOx) displaying a charac-
of GO, value of high resistance/OFF state kept on teristic bipolar resistive switching behavior with multi-
decreasing and hence, resulting in lesser gap between level resistive states and very high switching ratio of
the two resistive states. The device with the maximum 106. This was an extremely important study in the
Sci. Technol. Adv. Mater. 21 (2020) 167 M.M. REHMAN et al.

Table 3. Summary of GO/RGO-based RRAM devices.

Bottom electrode Top electrode Active layer Substrate Switching ratio Endurance Retention (s) References
Pt Cu GO Ti/SiO2/Si 20 102 104 [101]
ITO Al GO-PVK Glass 103 108 104 [146]
ITO Al GO Glass 103 102 109 [147]
ITO Al TPAPAM-GO Glass 103 – 104 [129]
Al Al GO PET 102 102 105 [148]
ITO Al AuNP-GO Glass 102 – 102 [122]
ITO Al RGO Glass 105 – – [121]
ITO Al GO Glass 103 102 – [149]
Pt Pt GO/PCMO SiO2/Si 102 – 104 [150]
ITO Ag GO-PI Glass 105 102 103 [151]
ITO Al PFCF-RGO Glass 104 108 104 [152]
ITO Al BCP-GO Glass 105 108 104 [153]
Pt Pt GO SiO2/Si 104 102 105 [123]
GO GO GO PET 102 103 103 [124]
Pt Al GO Si 104 102 103 [154]
ITO Ag GO PET 5 – 103 [155]
Ag Ag GO SiO2 10 – 103 [156]
PEDOT:PSS Au PVP/APTES/RGO/PVP/pentacene PES – 103 107 [131]
ITO Al GO PET 102 102 104 [149]
ITO Au GO – 102 – 103 [157]
ITO Al RGO SiO2 15 – 103 [158]
ITO Al GO-PANI Glass 103 108 104 [159]
ITO Al TiO2/GO Glass 102 105 103 [160]
ITO ITO RGO Glass 103 105 107 [127]
Ag Au RGO PET 104 102 105 [125]
Al Al RGO Glass 10 102 106 [126]
ITO Ag GO – 105 – 103 [161]
ITO Al GO-CISQDs Glass 104 105 – [162]
ITO Al GO-ZnO PET 102 102 104 [163]
Ta Pt GO/Ta2O5-x Si/SiO2 102 102 – [164]
Al Al ZnO/RGO SiO2 10 103 105 [132]
ITO Al RGO-PVA-AuNPs PET 103 50 103 [165]
ITO Al GO-AuNPs Glass 108 – 104 [166]
ITO Au GO Glass 10 102 – [167]
Al Al PS/GO/PS Si – 106 – [168]
ITO Al PVDF-GO Glass – – – [169]
ITO ITO GO PES 10 – 105 [170]
ITO Al GO-AuNPs Glass 105 102 104 [171]
Al Al GO SiO2 – – – [172]
Al Al GO Glass 102 102 104 [133]
ITO Al PVK:GONPs PET 102 – – [173]
Ti/Pt Ti/Pt GO Si/SiO2 103 104 105 [174]
ITO Al PMMA/GO Glass 103 104 105 [175]
ITO Al PTPEB-g-RGO Glass 104 102 104 [134]
ITO Al PW/GO Glass 103 102 104 [42]
ITO Al PVK-AZO-GO Glass 104 106 104 [135]
FTO Ag GO/SrTiO3 Glass 15 104 104 [137]
Pt Pt RGO Si/SiO2 105 – – [138]
Au Au PVA/GO-PVA/PVA Glass 104 102 103 [55]
ITO Ni PMMA:GO Glass 103 102 106 [144]
ITO Al PI-GO Glass 108 102 103 [136]
ITO Al OCQDs-GO – 102 102 105 [139]
ITO Al/Au GO Glass 104 – – [141]
ITO Al GO Glass – – – [142]
ITO ITO EVA/RGO PEN 105 – 104 [143]

context that the operating voltage (±0.2 V) and switch- monolayer of MoS2. The distinctive factor of these
ing ratio (106) of this MoS2-based memory device were devices was that it displayed negative differential resis-
far superior to several other devices based on either tance (NDR) behavior. The threshold voltage of these
graphene or MoS2. Moreover, the fabrication technol- devices was tunable from 3.5 V to 8 V with a variation
ogy was completely based on printing techniques that in the applied voltage at the gate terminal. This dis-
made the device fabrication extremely simple, quick tinctive structural change provided a new opportunity
and cost-effective. A schematic diagram of this device for designing bipolar and complementary switching
along with its electrical characteristic curves are illu- devices.
strated in Figure 11(a). Cheng et al. [181] explored a unique aspect of MoS2
Sangwan et al. [180] reported the bipolar resistive sheets by studying the effect of varying MoS2 phase on
switching effect in the monolayer of MoS2 with tun- its electrical properties. They proved that the 2 H phase
able characteristics of gate-controlled device in a field of bulk MoS2 had ohmic behavior while the 1 T phase of
effect geometry. The memory phenomena in this exfoliated MoS2 sheets exhibited a characteristic bipolar
device were mediated by the grain boundaries in the resistive switching effect when sandwiched between two
Sci. Technol. Adv. Mater. 21 (2020) 168 M.M. REHMAN et al.

Figure 11. (a) I–V characteristic curve and schematic diagram (inset) of the RGO/MoS2/ITO memory device along with its retention
graph [177]. (b) Schematic diagram illustrating the whole manufacturing process of as fabricated memory device on a flexible PET
substrate. The images enclosed in the red dotted block exhibit each step of the fabrication process while the images outside the
block correspond to the resulting device after each step. The bottom left corner exhibit layer by layer sandwiched structure of
a single memory cell in bend state with PET/Ag/MoS2- PVA/Ag configuration [8].

Ag electrodes. This device showed the least operating Liu et al. and Cheng et al. with an important additional
voltage value of ±0.2 V that was earlier achieved by feature that these devices were fabricated on a flexible
Bessonov et al. as well. A similar study by Zhang et al. PET substrate. Zhang et al. concluded that 1 T-MoS2
[182] was reported shortly after the work of Cheng et al. and 2 H-MoS2 devices exhibited two different types of
in which they also studied the effect of 1 T and 2 H NVM behaviors including WORM and rewritable
phase of MoS2 with a difference that instead of using memory, respectively. Wang et al. [183] proposed
MoS2 alone, they used a hybrid nanocomposite of MoS2 a photoresistive bistable memory device based on
-PVP that was earlier reported by Liu et al. This work MoS2 nanospheres. The control of this device was
can be thought as a combination of the study done by based on the polarization of nanospheres in light and
Sci. Technol. Adv. Mater. 21 (2020) 169 M.M. REHMAN et al.

dark conditions. Although the switching ratio of that materials including GQDS and CdSe QDs. Although
device was only ~10, multilevel resistive switching effect switching ratio displayed by these devices was average,
was achieved through varying illumination intensity of the SET (0.5 V) and RESET (−0.9 V) voltages were
white light. No resistive switching effect was observed at incredibly low that resulted in low power consump-
zero light intensity/dark conditions. Our research group tion. Later that year, Zhou et al. [186] reported the
also reported the characteristic bipolar resistive switch- characteristic bipolar resistive switching behavior in
ing behavior in the novel nanocomposite of MoS2 flakes the self-assembled three-dimensional microspheres of
interspersed in the organic polymer of PVA to form MoS2. The conduction mechanism in these devices
a hybrid nanocomposite [8]. The whole device was was associated with the formation and rupture of fila-
fabricated by a simple, cost-effective and state-of-the- ments. Guan et al. [187] reported the resistive switch-
art printing technology. Our device was flexible, all ing behavior in nanosheets of MoS2 blended with PVK
printed, rewritable, and nonvolatile. The distinctive polymer as reported earlier by Fan et al. with a few
feature of this device was its retention time and differences in its structure such as MoS2 nanosheets
mechanical robustness that were the highest as com- were thiol-modified, different top electrode, and an
pared to other MoS2 based memory devices. Our device extra layer of PEDOT:PSS polymer. Xia et al. [188]
showed a retention time of 105 s and mechanical robust- have explored an important aspect of bipolar resistive
ness of 1500 bending cycles without any significant switching behavior in the thin film of MoS2 by verify-
degradation in its characteristics. That device was bent ing the effect of changing top electrode material. This
with various diameters ranging from 50 mm to 2 mm. study depicted that Ag is the highly suitable material
Furthermore, the electrode patterning technique of to be used as the top electrode for MoS2 based memory
reverse offset for the fabrication of high-resolution devices as compared to Ti, Cu or Al due to the easy
RRAM devices was also reported for the first time. formation of Ag filaments through the redox process
The characteristics of this device were far superior to taking place inside MoS2 thin film. The easy oxidation
most of its predecessors either based on hybrid nano- of Ag in MoS2 thin film makes it a preferred choice to
composites or MoS2 RRAM devices. Mechanical flex- be used as a top electrode. Bhattacharjee et al. [189]
ibility of either MoS2 or PVA-based RRAM device was have reported the memory effect in the active layer of
also never reported before this work. The complete MoS2-PMMA nanocomposite that was a replication of
fabrication process along with a schematic diagram is the work done by our group earlier with the difference
shown in Figure 11(b). that the top and bottom electrodes were changed. The
Fan et al. [184] have reported a similar resistive change in electrode materials had a considerable effect
switching behavior in the hybrid nanocomposite of on the resistive switching characteristics of this already
MoS2 and poly(N-vinylcarbazole) with a configuration explored active layer as this time it exhibited multi-
of Au/MoS2-PVK/ITO. Although the memory beha- level resistive switching behavior instead of typical
vior exhibited by this device was neither poor nor bipolar behavior.
impressive, this was the first of its kind RRAM device Li et al. [190] investigated the switching behavior in
as PVK was covalently bonded with MoS2 nanosheets a few layer MoS2-based device in the form of a lateral/
to exhibit rewritable memory behavior for the first planar structure instead of the conventional sandwiched
time. The effect of varying the thickness of MoS2- structure. The main aim of this work was to suggest
PVK active layer on the resistive switching character- a method of emulating human neural network. They
istics was also explored and it was deduced that the figured out two switching modes including conductance
switching ratio kept increasing with increasing film mediated and rectification mediated, respectively.
thickness with no prominent change on the threshold They found out the transition from analog pulse-
voltages which normally is not the case. Fan et al. also programmed behavior in switching to the quasi binary
signified that annealing of MoS2-PVK active layer has memory on the application of electrical stresses for the
a major effect on the resistive switching characteristics as fabricated MoS2 memory device. They successfully
of this device including a drastic increase in the fabricated and coupled two MoS2 based memory cells
switching ratio (from 70 to 104) and a considerable and illustrated the potential in these devices to mimic
decrease in the SET and RESET voltages to reduce the human brain. Chen et al. [191] proposed a memory
power consumption. Wang et al. [185] also studied cell with double sandwiched layer for the application of
the memory behavior in the hybrid nanocomposite of emulating neural network to mimic the human brain. In
MoS2 and PMMA sandwiched between two layers of this device, MoS2 was used as a separating and blocking
PMMA polymer. This study was unique in a way that layer between two switching layers of transition metal
the MoS2 was used in the form of quantum dots for the oxides of HfO2 and TiO2. Perfect imitation of long-term
first time and their potential as a memory storage penetration (LTP) and short-term plasticity (STP) is
material was explored. These devices exhibited the vital for successfully mimicking the neural network of
lowest power consumption of ~2 mW when compared human brain. The success of this device was that it
with other RRAM devices based on QDs of other managed to emulate both LTP and STP due to its
Sci. Technol. Adv. Mater. 21 (2020) 170 M.M. REHMAN et al.

volatile memory behavior owing to the role of sand- its potential use in multifunctional electronic eyes and
wiched MoS2 layer that successfully blocked oxygen robotics. The summary of all the above discussed MoS2
ions. Hou et al. [192] explored the resistive switching based RRAM devices along with their typical character-
effect in a few layer MoS2-based RRAM device and istics are listed in Table 4.
simplified the three-terminal device to a two-terminal
device for the ease of fabrication and simple structure.
4. Role of other 2D materials in RRAM devices
There proposed device was an important step forward
to enhance the memory density. Prakash et al. [48] and MXenes are another important class of 2D materials
Kumar et al. [193] reported the resistive switching effect being discovered whose resistive switching behavior
in MoS2 based memory device with an interesting was first explored by Yan et al. [203] when they used
choice of bottom electrode as they used tungsten nitride a single nanorod of MoSe2 to connect a Ag anode with
(W2 N) and Ni-Mn-In ferromagnetic shape memory an Au cathode in a planar structure device. Although
alloy, respectively, instead of conventional electrode the switching ratio achieved by this group was only ~50,
like Pt, Au, ITO, etc. The obtained results of switching it introduced a new potential candidate for the active
ratio, endurance and retention were of average value but layer material in the future flexible nonvolatile memory
this bottom electrode was a novelty in this work. Kim devices. Inspired by these results, Han et al. [204] stu-
et al. [194] reported a breakthrough research by using died the memory effect in the nano-islands of MoSe2
the memory behavior of MoS2. They used its character- array embedded in a TiO2 matrix. The resistive switch-
istics to fabricate extremely low power RF switches for ing characteristic of this RRAM device exhibited stable
the next generation internet of things (IoT) and recon- memory behavior in both illuminated and dark condi-
figurable communication systems. A schematic diagram tions. The device exhibited photo-controlled memory
of their proposed device with the characterization of effect with the characteristic bipolar resistive switching
used materials is shown in Figure 12(d). Simplified behavior and its switching ratio was enhanced with
illustration of device structure and the signal transmis- increasing light intensity. Another study with a similar
sion of the RF switches are also presented. Qiu et al. approach was published by Zhang et al. [205] and they
[195] have recently proposed an important device struc- reported the memory effect in the nano-islands of
ture of RRAM device in which they have managed to MoSe2 grown through exactly the similar technique
tune the characteristics of resistive switching between on the same substrate with the only difference that these
bipolar, rectification and complementary resistive MoSe2 nano-islands were not embedded in TiO2 arrays.
switching by optimizing the interfacial layer of MoS2. This study also did not include the effect of photolumi-
The role of MoS2 in this device was to provide control- nescence on the resistive switching characteristics of
lable switching by confining electrons in it. Schematic MoSe2. The results of this device were extremely poor
diagram of the RRAM device with configuration of IrOx with a switching ratio of only ~12. Zhou et al. [206]
/Al2O3/TaOx/MoS2/TiN for the detection of ascorbic reported the RRAM behavior of MoSe2 doped ultra-
acid is shown in Figure 12(c). Das et al. [196] also long Se microwires. Although the switching ratio (~102)
reported a multilayer and multifunctional memory of this MoSe2 based memory device was much better
device based on MoS2 with the ability to store the binary than the one reported by Zhang et al., both the switch-
data along with sensing the visible light. They grew the ing ratio and retention time were insufficient. Recently,
layer of MoS2 nanoparticles on the surface of bottom Han et al. [207] have again reported the resistive switch-
electrode through heating. The particle size of as synthe- ing behavior in the hexagonal MoSe2 nanosheets. Han
sized MoS2 nanoparticles was in the range of 20 nm to et al. explored the nonvolatile memory behavior in the
50 nm. The device showed stable and long retaining nanosheets of hexagonal MoSe2 due to the formation of
stable results with multilevel storage and light-sensing Ag filaments. Unlike the typical sandwich structure of
ability with a high speed. Feng et al. [197] reported RRAM devices, this was based on a planar structure
a fully printed MoS2 based RRAM device on a flexible with Ag acting as the anode and cathode. The obtained
substrate whose schematic diagram and optical image results were average with an electrical endurance of 100
are illustrated in Figure 12(a). They achieved a high cycles and a switching ratio of 70. Most recently, Li et al.
switching ratio of 107 with a distinctive feature of both [208] have reported an extremely important study in
volatile and nonvolatile memory behavior in a single which they have explored the effect of temperature on
functional layer. Zhai et al. [198] have recently reported resistive switching and magnetic properties of MoSe2. It
another MoS2 based memory device whose complete was deduced from the obtained results that magnetism
fabrication process is illustrated in Figure 12(b). The of MoSe2 nanosheets had an inverse relation with tem-
distinctive feature of this device was its dual application, perature while their memory behavior was regulated by
i.e. as a memory device and infrared sensor. The func- temperature. The switching voltage of these devices
tional layer was based on the heterostructure of MoS2 were like those as reported by Han et al.; however, it
nanosheets with upconversion nanoparticles (UCNPs) kept on decreasing with increasing temperature, i.e.
to form MoS2–NaYF4:Yb3+, Er3+. This novel device has power losses in MoSe2 memory device can be decreased
Sci. Technol. Adv. Mater. 21 (2020) 171 M.M. REHMAN et al.

Figure 12. (a) Schematic structure of Ag/MoS2/Ag switch [181]. (b) Schematic illustration of the MoS2–UCNPs-based RRAM device
and main fabrication process [198]. (c) Schematic view of IrOx/Al2O3/TaOx/MoS2/TiN RRAM structure for detection of ascorbic acid
[195]. (d) Device schematics and images with material characterization. Simplified illustration of the signal transmission and device
structure of the RF switches based on monolayer MoS2. Zoomed-in plan view SEM image of a MoS2 RF switch with lateral area of
1 × 1 μm2. Scale bar, 2 μm. The dashed box in b marks the area covered with MoS2. The inset is a top-view optical image of
a fabricated MoS2 RF switch with Au electrodes. Scale bar, 50 μm [194].

by increasing temperature. However, the switching device based on molybdenum ditelluride (MoTe2).
ratio decreases with increasing temperature. A sche- They achieved a fast switching speed on less than 5 ns
matic diagram of their device along with its I–V char- with a better control over switching properties.
acteristic curves is illustrated in Figure 13(a). Zhang The first report on the resistive switching behavior of
et al. [51] reported the ultra-fast multi-level resistive hBN-based RRAM device was presented by Jain et al.
switching effect with a synaptic behavior in the RRAM [209]. They explored that the highly crystalline thin film
Sci. Technol. Adv. Mater. 21 (2020) 172 M.M. REHMAN et al.

Table 4. Summary of MoS2-based RRAM devices.

Bottom electrode Top electrode Active layer Substrate Switching ratio Endurance Retention (s) References
RGO Al MoS2-PVP PET 102 – – [176]
RGO ITO MoS2 SiO2 104 – 103 [177]
Ag Ag MoS2/MoOx PEN 106 – 103 [179]
FTO Ag MoS2 Glass 103 102 – [199]
Au Au MoS2 Si/SiO2 103 – – [180]
Au Au MoS2 Si 10 103 – [183]
Ag Ag MoS2 Glass – 102 – [181]
ITO Al MoS2-PVP PET 103 104 102 [182]
Ag Ag MoS2-PVA PET 102 103 105 [200]
ITO Ag MoS2 Glass 104 102 103 [186]
ITO Cu MoS2-PMMA PET 103 105 105 [189]
FTO Au PMMA/PMMA-MoS2QDs/PMMA Glass 102 102 104 [185]
ITO Au MoS2/PVK – 102 102 104 [184]
ITO Al MoS2-PVK/ Glass 102 102 – [187]
PEDOT:PSS
Ti Ti/Al/Cu/Ag MoS2 Si 10 – – [188]
TiN Au HfOx/MoS2/TiOx – – – – [191]
ITO Al MoS2 Glass 102 104 107 [196]
Ag Ag MoS2 Kapton 108 105 [197]
Cr/Au Cr/Au HfO2/Al2O3/ Si 104 102 103 [192]
HfO2/MoS2
W2 N Cu MoS2 Si 103 103 103 [48]
Ti Ag ZnO/MoS2 Ti 2 102 – [201]
Ni-Mn-In Cu MoS2 Si 102 102 103 [193]
Ti/Au Ti/Au MoS2 Si – – – [190]
TiN IrOx/Pt/Ru Al2O3/TaOx/MoS2 – 105 103 – [195]
Au Au MoS2 Glass 103 20 104 [194]
ITO ITO HfOx-MoS2-PdNPs Glass 103 102 104 [39]
Ti Au MoS2/PbS Si 102 103 104 [202]
ITO Al MoS2–UCNPs Glass 104 102 104 [198]

of hBN insulator exhibited unipolar resistive switching based on formation and rupture of metallic filaments
behavior as it was not dependent on the polarity of through studying the effect of temperature and device
applied bias. Although Jain et al. showed the potential size on the bistable resistive states (HRS/LRS). The step
of data storage in hBN, they did not study its switching by step fabrication process of our proposed memory
properties in depth and this extremely important 2D device is shown in Figure 13(c). Recently, in the year
material was left high and dry for a couple of years 2019, Jing et al. [211] have reported the resistive switch-
until Qian et al. [210] reported the thorough study on ing-based nonvolatile memory behavior in hBN-based
the resistive switching properties of hBN by fabricating RRAM device deposited through CVD. This was an
the simplest structure of RRAM device in which thin film important study as it proposed a transfer-free fabrication
of hBN was sandwiched between two metallic electrodes. technique of multilayer hBN thin film on a nickel-coated
The results illustrated in this study were important as wafers of silicon. The conventional techniques were
they showed the potential of dielectric hBN as a flexible expensive, slow and cracks used to appear in the thin
memory material. They also explored the effect of chan- film during the transfer process that could directly alter
ging top electrode on the switching properties of hBN the electrical characteristics of the memory device. They
and evaluated the results of its mechanical robustness by attributed the resistive switching effect of their memory
bending it up to a radius of 7 mm. The memory behavior device to the percolation paths across the hBN stacks and
illustrated by this dielectric 2D material was comparable thin layer of TiBN formed at the interface Ti/hBN.
to its predecessor 2D materials such as graphene and Tungsten disulfide (WS2) is yet another 2D material
MoS2. A schematic diagram of their RRAM device along with promising electrical and mechanical properties
with its characteristic I–V curve is shown in Figure 13(b). whose resistive switching properties were not explored
Inspired by the promising results of Qian et al. our until recently our group published the results exhibit-
research group formed a hybrid nanocomposite of hBN ing the memory behavior of WS2. The switching char-
with PVA polymer sandwiched between ITO and Ag on acteristics illustrated by WS2 were far better than
a flexible PET substrate to provide highly stable, scalable many other earlier reported single 2D-material-based
and reliable characteristic bipolar resistive switching RRAM devices including electrical endurance and
behavior [9]. Our device was fabricated through simple mechanical robustness. This study will encourage
and cost-effective all printed technology. The obtained other researchers to explore the switching character-
results of this device were superior to many other RRAM istics of WS2 based nanocomposites to further
devices based on nanocomposites of PVA polymer improve its switching ratio and power consumption.
including its mechanical robustness, switching ratio, These devices showed impressive retention time of
retention time, power consumption and electrical endur- 105 s as they were encapsulated with atomically thin
ance. The conduction mechanism was proved to be film of aluminum oxide (Al2O3) to enhance their
Sci. Technol. Adv. Mater. 21 (2020) 173 M.M. REHMAN et al.

Figure 13. (a) The current–voltage (I–V) curves of Ag/MoSe2/Ti structure at 300 K, 400 K and 500 K, respectively; the inset shows
the device structure [208]. (b) The switching characteristics for the hBN memory device after electroforming process with
Schematic of the Ag/hBN/Cu foil on PET substrate device arrays [210]. (c) Schematic diagram of entire device fabrication through
all-printing technology followed by encapsulation of atomically thin film of Al2O3 to enhance its lifetime [213].

lifetime. Das et al. [212] were the second group in line based on other 2D materials (hBN, MoSe2, MoTe2,
after our team to propose a memory device based on WS2, and WSe2) along with their typical characteris-
the active layer of WS2. The thin film of WS2 was tics are listed in Table 5.
grown through the CVD. They achieved a switching
ratio of 103 with 100 endurance cycles. Recently Li
et al. [53] have proposed a completely printed few
5. Multi-2D-materials-based RRAM devices
layer WSe2-based RRAM device fabricated through
aerosol jet printing technique in the year 2019. This Several studies have been carried out on exploring the
device had an interesting characteristic as it showed resistive switching characteristics of hybrid nanocom-
both volatile and non-volatile unipolar memory beha- posites, i.e. blending a 2D material with other materi-
vior with a reasonable retention time of 2.5 h. The als such as polymers however only a few reports are
device was fabricated on a flexible Kapton tape and available in which a blend of two or more 2D materials
gave stable results up to a bending radius of 5 cm. The is used as the functional layer of RRAM device. The
summary of all the above-discussed RRAM devices first study on the nanocomposite of two 2D materials
Sci. Technol. Adv. Mater. 21 (2020) 174 M.M. REHMAN et al.

Table 5. Summary of other 2D-Materials (hBN, MoSe2, MoTe2, WS2 and WSe2)-based RRAM devices.
Bottom electrode Top electrode Active layer Substrate Switching ratio Endurance Retention (s) References
ITO Ag hBN/-PVOH PET 102 103 – [9]
ITO Ag/Au/Cu hBN PET 102 102 103 [210]
Ag Ag hBN/GQDs PET 103 103 105 [40]
Ag Au MoSe2 Si/SiO2 50 102 – [203]
FTO Ag MoSe2-TiO2 Glass 102 50 – [204]
FTO Ag MoSe2 Glass 5 50 – [205]
Ag Ag MoSe2 SiO2 102 102 103 [206]
Ag Ag MoSe2 Si 70 102 – [207]
Ti Ag MoSe2 Glass 10 – – [208]
Ti/Au Ti/Ni MoTe2 SiO2 10 102 105 [51]
Pt/Ti Al WS2 Si/SiO2 103 102 – [212]
Ag Ag WSe2 Kapton 103 – 104 [53]
Ag Ag WS2 PET 103 102 105 [213]

was performed by Yin et al. [214] who used a blend of Recently, many research groups tried to fabricate
MoS2 and graphene oxide as the sandwiched layer a memory device by using two or more than to 2D
between two metallic electrodes. Although the materials in a single RRAM device. These groups
RRAM devices based on graphene oxide showed include Zhu et al., Choudhary et al., Wu et al., and
impressive results, the motivation of this study was Yalagala et al. Zhu et al. [43] reported an entirely 2D
to lower the values of SET and RESET voltages of GO- material RRAM device with three stable resistive
based memory devices by forming a blend with MoS2 states that could be used for multiple storage of
in order to reduce power consumption. The obtained data. They sandwiched a multilayer thin film of
results were encouraging as the conductivity of MoS2- hBN between multilayer thin films of graphene with
graphene oxide thin film was increased that resulted in a device configuration of Au/Ti/G/hBN/G/Au and
decreasing the threshold voltage up to ≤1.5 V with device area of 5 μm × 5 μm. Three instead of two
a stable switching ratio of 102. resistive states were realized by limiting the magni-
Shinde et al. [215] fabricated RRAM device with tude of compliance current. Furthermore, all the 2D
a different structure in which they used bilayer func- materials (two layers of graphene and a single layer of
tional layer with the first layer of graphene and hBN) used in this work were deposited through CVD
the second layer of an organic polymer (PMMA) instead of using the conventional technique of
blended with MoS2. This pair of MoS2 and graphene mechanical exfoliation. A schematic diagram along
exhibited ten times better switching ratio as compared with the electrical characterization results are shown
to the MoS2-graphene oxide reported earlier by Yin in Figure 14(b). Choudhary et al. [47] reported
et al. Shin et al. [216] published an article to report the a RRAM device based on the combination of GO and
nonvolatile memory behavior in the functional thin MoS2 nanocomposite. The effect of changing bottom
film of MoS2 nanosheet embedded with graphene electrode on the switching characteristics of this
oxide. The obtained results were very important as device was also studied by using Al and ITO in two
the fabricated device displayed multilevel resistive different devices. Interestingly the device with the Al
switching with an added advantage of high-density bottom electrode illustrated superior RRAM beha-
data storage. Nanosheet of MoS2 was sandwiched vior as compared to the ITO with a ten times better
between two layers of graphene oxide to exhibit at OFF/ON ratio and low values of threshold voltages
least four resistive states and average switching ratio for SET/RESET states. The possible reason for this
of 102. This work also verified that embedding MoS2 difference can be attributed to the higher concentra-
between two dielectric layers like graphene oxide tion of oxygen vacancies at the interface of active
makes it an excellent charge storing material. Owing layer and Al electrode that is not present at the inter-
to the huge potential of making the nanocomposites of face of this device with ITO as the bottom electrode.
2D materials, our research group has recently explored Wu et al. [217] also proposed a similar device as that
the resistive switching behavior in the novel nanocom- of Choudhary et al. with the prominent features of
posite of hBN-GQDs. The device was all-printed and multilevel storage, low operating voltage and high
had a planar structure with a protective layer of atom- flexibility. They used RGO instead of GO as the
ically thin film of Al2O3 deposited by spatial atmo- composite material with MoS2 and their choice of
spheric atomic layer deposition (SAALD) system. This electrodes combination was also distinct. They also
device was fabricated on a flexible PET substrate with achieved multiple resistive states in their fabricated
impressive electrical and mechanical properties such device by controlling the magnitude of compliance
as switching ratio (103), electrical endurance (103 bias- current. This device showed better electrical and
ing cycles), retention time (105 s) and mechanical mechanical characteristics of a nonvolatile RRAM
robustness (2000 cycles) without any prominent device. The optical image, schematic diagram and
decay in either resistive state (LRS/HRS). I–V characteristic curve of the proposed RRAM
Sci. Technol. Adv. Mater. 21 (2020) 175 M.M. REHMAN et al.

Figure 14. (a) Photo of the Ti/MoS2-rGO/ITO flexible memory device and schematic of the memory unit of the device. Typical I-V
curve of the Ti/MoS2- rGO/ITO device plotted on a semilogarithmic scale; the inset displays the typical I-V curve of the Ti/ rGO/ITO
device[45].(b) Device structure and forming process analysis. Optical microscope image of a matrix of deviceswith cross-point
structure, and its cross-sectional schematic with Au/Ti/G/h-BN/G/Au/Ti/SiO2/Sistructure from top to bottom. Forming I-V curves of
one fresh Au/Ti/G/h-BN/G/Au memristor (black) and a control Au/Ti/G/G/Au structure (red); the results indicate that the dielectric
breakdown in related to the h- BN insulating stack. Statistical analysis of VBD and IBD shows an increasing VBD and decreasing IBD
with smaller device area[46].

device are shown in Figure 14(a). Yalagala et al. [218] 6. Conduction mechanism of 2D-materials-
proposed a memory device based on the active layer based RRAMs
of graphene and MoS2 with Ag and Cu as the top and
RRAMs change their states between HRS and LRS on
bottom electrodes, respectively. They experimented
the application of externally applied electric field due
the fabrication of their device on a flexible paper
to change in conductivity of a material. Conduction
substrate to make their device cost effective and easily
mechanisms of 2D-materials-based RRAMs can be
disposable. The summary of all the above-discussed
broadly divided into two categories including interface
RRAM devices based on other composite 2D materi-
of electrode-dielectric (involving the formation of
als along with their typical characteristics are listed in
a filament and oxygen vacancies) and the other one
Table 6.

Table 6. Summary of multi-2D-materials-based RRAM devices.

Bottom electrode Top electrode Active layer Substrate Switching ratio Endurance Retention (s) References
ITO Al MoS2-GO Glass 102 – – [214]
SiO2 Au Graphene/PMMA-MoS2 Si 103 – – [215]
Al Au GO/MoS2/GO SiO2 102 102 108 [40]
Au Au Ti/Graphene/hBN/Graphene Si/SiO2 104 102 – [43]
Al ITO MoS2-GO Si 102 104 104 [47]
ITO Ti MoS2-RGO PET 10 102 104 [217]
Ag Cu Graphene-MoS2 Paper 104 102 104 [218]
Sci. Technol. Adv. Mater. 21 (2020) 176 M.M. REHMAN et al.

dependent upon the active material itself only (SCLC, 7. Challenges and future prospects
ohmic conduction, Poole-Frenkel emission, hopping
A few of the RRAM devices based on graphene regis-
conduction, and redox mechanism). Determining the
tered record-breaking endurance value while other
exact switching mechanism of RRAMs has never been
2D-material-based devices illustrated good switching
an easy task but owing to the advancements in tech-
speeds and low operating voltages (<1 V). These 2D-
nology, more insights into this difficult yet important
materials-based RRAM devices have shown a lot of
problem are being explored. SCLC is the highly
promise in only a decade of research behind them as
reported conduction mechanism in 2D-materials-
compared to more than six decades of research work
based RRAMs in which charge carriers follow Ohm’s
on TMO-based RRAM devices that is an encouraging
law or thermionic emission at low applied voltage but
sign for the researchers to keep working in this area.
when the magnitude of applied voltage reaches the
Among the other factors that need to be improved in
threshold voltage, more charge carriers are injected
near future for the development of 2D materials-based
and conduction shifts from ohmic to SCL. The elec-
RRAMs, the fabrication technology is one of the key
trode material also plays a vital role in the conduction
areas. For instance, one of the best endurance values
mechanism of RRAMs due to filament formation/
achieved in a non-carbon 2D material-based RRAM
migration of oxygen vacancies. Filament formation
device (103 cycles in Ag/MoS2/Ag) was fabricated by
occurs when the atoms of top electrode begin to dif-
using Ag paint as the top electrode and Ag foil as
fuse into the active layer itself whereas the SCLC
bottom electrode with a cell size of ≈ 0.1 mm2. These
mechanism is dependent upon the electrical properties
fabrication methods are not suitable for industry
(trap spacing, energy level, trap density, drift mobility,
therefore, the future work in 2D-material-based
etc.) of 2D material. Growth of conductive filaments is
RRAMs should focus on the development of new
due to the accumulation of defects such as metallic
fabrication techniques that are highly compatible
cations and oxygen vacancies that causes the device to
with industry. Annealing treatments after transfer
operate in SET state while the formed filament is
process to eradicate remains of polymer may also be
usually deformed through the thermal or electric
an option. Consequently, transfer process should be
effect. The conduction mechanism based on filament
eluded whenever possible, not only because of the
formation and rupture can be divided into three cate-
problems of device performance, but also because it
gories, namely, valence change mechanism (VCM),
elongates the fabrication process hence resulting in
electrochemical metallization mechanism (ECM) and
higher fabrication cost. The best result would be to
thermochemical mechanism (TCM) whose details
mature transfer-free methods, but the direct growth of
have been discussed by Li et al. [219]. Diameter
graphene-related materials on transition metal oxides
(size) and quantity of conductive filaments determine
is a long-term objective that will take further time for
the performance of RRAM device. Uncontrolled fila-
realization. The use of insulating layered materials as
ment growth can result in poor performance such as
resistive switching medium is desired because, primar-
low stability, low endurance and short lifetime. Efforts
ily, they do not need to be transferred, and, secondly,
must be done to control the growth and rupture of
the lack of cracks can be verified by the observation of
filament growth. This process can somehow be con-
a forming set process. Recent research works on
trolled through doping the resistive switching layer of
RRAM devices based on layered materials have
2D material and by inserting another layer.
reported CVD-grown hBN that does not require any
The switching mechanism of 2D-materials-based
transfer step with a deficiency that they also used
RRAMs can also be further classified in terms of
metallic foils. When using hBN as a resistive switching
polarity of applied voltage, i.e. bipolar switching and
medium, it becomes possible to avoid the standard
unipolar switching. In bipolar type switching, SET and
transfer process, and the catalyst substrate for the
RESET process occur at two different polarities
CVD growth can be used as a bottom electrode. The
whereas in unipolar switching, both SET and RESET
direct growth of graphene-related materials by CVD
process can occur at either (positive or negative) pola-
on flat wafers coated with metal is highly desirable.
rities. In unipolar mechanism, the switching between
A different option is to use LPE graphene-related
bistable resistive states occurs due to the magnitude of
materials insulators that can be spin-coated on ran-
applied voltage. Bipolar and unipolar switching
dom substrates, but they might present inconsistency,
mechanism can coexist in a single RRAM device,
given their large unevenness (classically ≈20 nm),
such RRAMs are said to have nonpolar switching
much more than flat graphene-related materials pre-
mechanism. From the above discussion, it can be
pared by CVD. The use of coating techniques that
deduced that the conduction mechanism of a RRAM
could decrease the irregularities below 1 nm is com-
device based on 2D materials is a very sensitive topic
pulsory. Note that the irregularities of transition metal
that can be affected by the choice of electrode material,
oxides for RRAMs (typically grown by ALD) are
fabrication technology, film thickness of the 2D mate-
≈0.2 nm. Future work on RRAM devices based on
rial, etc.
Sci. Technol. Adv. Mater. 21 (2020) 177 M.M. REHMAN et al.

2D materials should be aimed at reducing the device The fifth aspect that can be further explored for
size (<1 μm2), emphasis on representing device per- improving existing 2D-materials-based RRAMs is to
formance (i.e. retention, endurance, power consump- focus on the characterization techniques to character-
tion, and switching time) above the requirements of ize nanoscale surfaces, such as topographic AFM
NVM technology and contain analyses of variability maps. Researchers have successfully reported the for-
and reliability. mation and rupture of nano-filaments from one elec-
The second challenge related to transferred single- trode to the other electrode through the active layer
layer graphene as interface electrode RRAMs is their and verified their theoretical analysis by using high-
potential to be further reduced in size. The major resolution TEM images. The sixth contribution in the
disadvantage faced by these devices due to large size field of 2D-materials-based RRAMs can be the use of
is that several cracks begin to appear in single-layer atomistic modeling and simulations to explain the
graphene mostly after being transferred. These cracks tentative observations. Only a little work has been
in turn become the reason for the formation of con- done on simulating the results of 2D materials-based
ductive filaments in large size memory cells based on RRAMs. Software like COMSOL can be used for simu-
single-layer graphene that play undesired role in the lating the results, varying the values of different vari-
conduction mechanism. Since the retention and ables and defining the boundary conditions of 2D
endurance times are related to the conductive filament RRAMs. Simulating the results of such devices will
properties, the presence of cracks and their influence be helpful in predicting the potential of certain mate-
cannot be neglected. Future endeavors should be aim- rials and device structure before directly starting
ing to make sure that no cracks are present in the experiments which will in turn save time and materi-
transferred single-layer graphene electrodes. Other als. Furthermore, researchers can easily vary the values
than reducing the cell size, another alternative is the of different parameters like device size, film thickness,
use of MLG, which has fewer cracks and is more etc. to observe their effect on the switching character-
resilient towards mechanical fractures during transfer istics of the device. The seventh area of research
process. Several graphene-related-materials-based related to 2D-materials-based RRAMs is to fabricate
RRAM devices have been fabricated by polymer- them on flexible substrates due to their characteristic
scaffold-assisted transfer method that did not assess of immense mechanical endurance. This distinctive
the occurrence of residues on the surface of graphene- characteristic of 2D-materials-based RRAM devices
related materials after removing the polymer. These will pave a new way for the development of next-
might diminish the size of device, given their insulat- generation flexible NVM RRAM device. A summary
ing nature and higher thickness (>10 nm). This pro- of flexible 2D-materials-based RRAM devices has been
cess is arbitrary, but can be concentrated by using compiled in Table 7.
superior cleaning techniques, that might result in The future directions for the researchers working in
device-to-device inconsistency. The third important 2D materials-based RRAM devices can be summarized
challenge being faced by the 2D-materials-based as below:
RRAMs lies in the fact that how excellent device per-
formance can be achieved through engineering device (1) Fabrication technology for the development of
structure, i.e. either by varying the electrode or the 2D materials-based RRAM device will see
functional layer. Various device structures have major improvements, especially in the synthesis
already been explored including the planar structure, of 2D materials and eliminationof the transfer-
sandwich structure, bilayer RRAM, trilayer RRAM, ring step involved in majority of such devices.
using two layers of different 2D materials, using 2D The fabrication technology should be quick,
material with a polymer, etc. This is still an open area cheap and have a high throughput and above
for the researcher working in the field of 2D-materials- all, it should be compatible with industry.
based RRAMs. (2) Reducing the number of cracks to enhance the
The fourth significant area open for the researchers electrical characteristics of 2D materials-based
to contribute is to work on further understanding the RRAM devices as these cracks are responsible
electrical switching behaviors and conduction mechan- for the formation of conductive filaments. The
ism of 2D-materials-based RRAMs that have been electrical characteristics such as endurance,
rarely explored. It is vital to carefully engineer the retention time and switching ratio can be
design of the as developed memory devices based on improved. There are two possible ways to
2D materials to attain their high-performance in future. reduce the number of cracks, i.e. by reducing
It should be well known that several aspects are impera- the device size and by using multilayers of 2D
tive in the realization of high-performance NVM materials instead of single layers.
devices based on 2D nanomaterials. For instance, the (3) Device structure of 2D materials-based RRAM
quantity of functional groups on 2D nanosheets (e.g. devices still needs to be finalized including the
GO) might affect the device performance. selection of number of layers, sandwich or
Sci. Technol. Adv. Mater. 21 (2020) 178 M.M. REHMAN et al.

Table 7. Summary of 2D-materials-based devices fabricated on a flexible substrate; hRGO and lRGO stand for highly reduced and
lightly reduced graphene oxide, respectively.
Device Substrate Radius Mechanical Electrical Ratio References
Al/PMMA/graphene/PMMA/ITO PET 10 mm – 105 107 [83]
SLG/Al2O3/HfOx/Al2O3/ITO PEN 5 mm – – – [90]
ITO/G-PVP/Ag PET 15 mm – 40 35 [120]
G/CNT/Al2O3 PET 8 mm 103 102 103 [60]
MLG/PMMA/Al PET 4.2 mm 104 – 106 [61]
G/P3BT:PMMA/Al PET 10 mm 107 107 106 [64]
G/TiO2/Pt PEN 10 mm 102 – 102 [65]
Al/GO/Al PES 9 mm 103 – 102 [148]
Al/GO/ITO PET 4 mm 102 102 103 [220]
hRGO/lRGO/hRGO PET 5 mm 103 – 103 [124]
ITO/GO/Ag PET – – – 5 [155]
PEDOT:PSS/PVP/APTES/RGO/ PES – – 102 – [131]
PVP/pentacene/Au
ITO/GO/Al PET – – 102 102 [149]
Ag/Al2O3/RGO/PMMA/Pentacene/Au PET 5 mm 102 102 104 [125]
ITO/ZnO/GO/Al PET 4 mm 103 102 102 [163]
Al/GO/Cu PES 4 mm – 90 104 [221]
ITO/Au NPs-RGO-PVA/Al PET – 50 – 103 [165]
ITO/GO/ITO PES – – – 10 [170]
ITO/PVK:GO/Al PET – – – 102 [173]
Pt/GO/Ti/Pt PEN 2 mm 103 102 104 [174]
ITO/hBN-PVOH/Ag PET 4 mm 102 103 102 [9]
Cu/hBN/Ag PET 7 mm 102 102 102 [210]
Ag/GQDshBN/Ag PET 2 mm 102 103 103 [40]
RGO/MoS2-PVP/Al PET – – – 102 [176]
Al/2 H-MoS2 -PVP/ITO PET – – 104 102 [182]
Ag/MoS2-PVA/Ag PET 2 mm 102 103 102 [8]
ITO/PMMA-MoS2/Cu PET – 102 105 103 [189]
Ag/WS2/Ag PET 5 mm 102 102 103 [213]
G/MAPbl3/Au PET 4 mm 103 102 50 [73]
ITO/MgO-PVP-G/Ag PET – 102 102 103 [44]
Ag/WSe2/Ag PEN 5 cm – – 103 [53]
ITO/MoS2-RGO/Ti PET 10 mm – 102 15 [217]
Cu/GO-MoS2/Ag Paper – 103 102 104 [218]

planar structure and pair of materials to be used 8. Conclusions

in those layers.
In this review paper, we have thoroughly compiled and
(4) Various conduction mechanisms of 2D-
discussed the advancements in the RRAM devices based
materials-based RRAM devices have been pro-
on 2D materials over the last decade and compared their
posed in the last decade; however, likewise
performance with each other as well as with RRAMs
other RRAMs, this particular aspect needs to
be further explored for the better understand- based on other materials. These 2D materials include
ing of this field. graphene, GO, hBN, MoS2, WS2, MoSe2, and composites
(5) Advancement in the characterization techni- of 2D materials. The contribution from various research-
ques to characterize nanoscale surfaces is ers working in this area has been thoroughly discussed by
needed. highlighting their strengths and weaknesses. After the in
(6) Simulations of 2D materials-based RRAM depth analysis of 2D-materials-based RRAMs over the
devices before experimentally fabricating last decade, we can conclude that the unique character-
them will help in saving materials, reducing istics of these ultrathin, layered materials have proved to
the cost of fabrication and predict the effect be helpful in fulfilling the highly desired requirements of
of varying various parameters such as device flexible, nonvolatile, reliable and transparent memory
size, material composition, environmental device of next generation. Though numerous RRAM
conditions, etc. devices based on 2D materials have been reported to
(7) One of the major advantages of 2D materi- date but still there is more room left for improvement
als-based RRAM devices over other RRAMs in order to finally commercialize these devices.
is the mechanical endurance of 2D materials Graphene is the most commonly used layered mate-
due to which they can be used for the fabri- rial in the fabrication of 2D-materials-based RRAM
cation of flexible devices. The future of elec- devices over the last decade. They have been presented
tronics is heading towards flexible devices so in the structure of RRAMs with the aims of i) increasing
researchers should try to find a way in which their performance as a nonvolatile memory (switching
flexibility can be introduced in all 2D mate- time, operating voltages, retention, endurance, power
rials-based RRAMs without compromising consumption) and ii) provide added abilities (transpar-
their characteristics. ency, heat dissipation, flexibility, chemical stability).
Sci. Technol. Adv. Mater. 21 (2020) 179 M.M. REHMAN et al.

Graphene can be used either as an electrode material in material have different memory characteristics.
RRAMs to provide transparency and flexibility or as an Another important parameter that has a huge
interface layer between two metallic electrodes to func- impact on the memory characteristics of such
tion as a resistive switching medium in order to dimin- devices is the number of layers of that particular
ish the cycle-to-cycle variability, by eluding atomic 2D material that has to be used as the active layer.
diffusion between insulator and electrode. This can Thus, organized research efforts are required to
decrease the power consumption due to its low contact improve the impact of all these factors on the
resistance as compared to the metallic electrodes. overall device performance, including the hybridi-
Moreover, it also suppresses the surface effects by elud- zation with other materials of the active layers,
ing photodesorption and chemisorption. RRAMs based material of electrodes, thickness of memory cell,
on graphene-related materials have displayed reprodu- surface functional groups, effect of environmental
cible bipolar and unipolar resistive switching behavior conditions (temperature and humidity), etc. It can
with low operating voltages <1 V, high switching ratio be concluded that graphene is the best 2D material
>105, and fast switching speed. In majority of reports, for RRAM devices that are meant to be used at
the switching mechanism is endorsed to the formation/ extremely high and low temperatures. Graphene is
disruption of conductive filaments in the resistive also the best choice among 2D materials that can
switching medium, and the atomic reorganizations in be used as an electrode of a RRAM device.
each state change are linked to the movement of inher- Graphene has a major drawback that it is mostly
ent species and diffusion of metallic ions from neighbor- produced by CVD; however, if the use of CVD is
ing layers, displaying parallelism with RRAMs based on to be avoided, hBN is the best choice among 2D
transition metal oxides. Graphene related materials have materials as the standard fabrication method can
also been embedded with nanoparticles, polymers, be avoided in this case. On the other hand, TMDs
quantum dots, and nanorods in order to increase the are useful if a flexible range of bandgaps and
performance (primarily endurance and retention). The tunable oxidation states are needed in RRAMs.
drawback of graphene is that it is typically formed by the Although the most commonly used TMD, i.e.
complicated process of CVD and inserted in RRAMs by MoS2 can not match the resistive switching char-
polymer-assisted transfer process whereas BP and GO acteristics of graphene MoS2 can be used by che-
are typically formed by relatively easier methods of spin mical functionalization or through dissolving it
coating and LPE on a conductive wafer that functions as with other polymers. Graphene-based RRAMs
a bottom electrode. should be avoided to reduce the cost and speed
The second most explored 2D material for the up the fabrication process. A few of the research
fabrication of RRAMs is MoS2 in various forms groups have also reported the optical behavior of
(flakes, nanoparticles, quantum dots, etc.). The 2D material-based RRAMs in which MoS 2 is
remaining 2D materials (hBN, WS2, MoSe2, a better choice as compared to graphene as it has
WSe2) based RRAM devices have also been a direct bandgap of about 2.5 eV. On the other
recently explored as the potential active layer of hand, graphene has a much higher thermal con-
the flexible nonvolatile RRAM. Based on the ductivity as compared to other layered materials
obtained results, it can be deduced that even by that makes it a suitable choice for temperature
using the identical active layer of 2D material, the RRAM devices whose conduction mechanism
memory characteristics such as switching ratio, mainly depends on energizing the electrons
threshold voltage, electrical endurance and reten- through heat energy. In order to achieve the best
tion time are dissimilar from each other which mechanical properties of a 2D material-based
implies that there must be other controlling para- RRAM, a heterostructure of graphene and MoS2
meters involved in affecting the device character- can be formed that will complement the mechan-
istics such as varying the thickness of functional ical constraints of each other as graphene has
thin films and electrodes. Furthermore, the switch- a large Young’s modulus and large yield stress
ing mechanism of a memory device can also be whereas MoS2 has a higher bending modulus.
controlled by altering the ratio of blended materi- It took more than five decades of extensive research
als in the active layers based on composites of 2D endeavors for TMO-based RRAMs to achieve high pro-
materials. For instance, an active layer formed by cessing speeds, decent endurance, low power consump-
blending the 2D material with a polymer is greatly tion, small size, high integration capacity and long data
affected by the concentration of 2D material. retention times; however, RRAMs based on 2D materi-
Moreover, the geometry/shape in which the same als have already come close to these devices within
2D material is used in the RRAM device also a single decade. Despite all determinations, nonvolatile
affects its properties so that nanorods, quantum memories technological requirements like long data
dots, flakes and nanoribbons of the same 2D retention >10 years and high endurance >109 cycles
Sci. Technol. Adv. Mater. 21 (2020) 180 M.M. REHMAN et al.

still stay a challenge. From the standpoint of flexibility, Dr. Jahan Zeb Gul received his PhD in
graphene-related materials-based RRAM devices can Mechatronics Engineering from Jeju
hold resistive switching behavior under more than National University, South Korea in
2019 with major research work in Bio-
>105 bending stresses with radius as low as to Inspired Soft Systems, Sensors, Bio-
few mm. Most studies based on resistive switching Robots, Actuators and Soft Active
devices fabricated by using 2D materials are focused Materials. He is currently serving as
on proof-of-concept illustrations using large area an “Assistant Professor” in the depart-
(>2000 μm2) devices, that makes it tough to infer to ment of Mechatronics & Biomedical
Engineering, AIR University, E9, Islamabad and member
real ultra-scaled RRAM devices but the remunerations
of SESL (Smart Engineering Systems Lab). For the past 5
of other 2D material properties (such as thermal heat years he is actively engaged in the field of “Bio-Inspired Soft
dissipation and high chemical stability) on the perfor- Systems for biomedical applications”. This include research
mance of RRAM devices have not been explored yet. publications, book chapters, industrial projects, and medical
Furthermore, research must also be done on the simu- prototypes.
lations of 2D-materials-based RRAM devices.

References
Acknowledgments
[1] Luo Y, Ghose S, Cai Y, et al. HeatWatch: improving
This work was supported by the Higher Education 3D NAND flash memory device reliability by exploit-
Commission (HEC) of Pakistan under SRGP grant (1982). ing self-recovery and temperature awareness. 2018
The authors are grateful to the support from Ghulam Ishaq IEEE International Symposium on High Performance
Khan Institute of Engineering Sciences & Technology Computer Architecture (HPCA); 2018 Feb. Vienna,
(Pakistan) and Jeju National University (South Korea) Austria: IEEE. p. 504–517
where this work has been done. [2] Xiang J, Li X, Ma Y, et al. Efficient flash memory devices
based on non-conjugated ferrocene-containing
copolymers. J Mater Chem C. 2018;6:11348–11355.
[3] Mondal S, Venkataraman V. Low temperature below
Disclosure Statement 200°C solution processed tunable flash memory
No potential conflict of interest was reported by the authors. device without tunneling and blocking layer. Nat
Commun. 2019;10:2143.
[4] HIDAKA H. Embedded flash memory for embedded
systems: technology, design for sub-systems, and
Funding innovations. ed H Hidaka. Cham: Springer
This work was supported by the Higher Education International Publishing; 2018.
Commission, Pakistan [1982]. [5] Aritome S. Scaling challenge of NAND flash memory
cells NAND flash memory technologies. Hoboken, NJ,
USA: John Wiley & Sons, Inc.; 2015. p. 129–193.
[6] Hui F, Grustan-Gutierrez E, Long S, et al. Graphene
Notes on contributors and related materials for resistive random access
memories. Adv Electron Mater. 2017;3:1600195.
Dr. Muhammad Muqeet Rehman
[7] Siddiqui GU, Rehman MM, Choi KH. Enhanced resis-
received his PhD degree in
tive switching in all-printed, hybrid and flexible mem-
Mechatronics Engineering from Jeju
ory device based on perovskite ZnSnO3 via PVOH
National University, South Korea in
polymer. Polymer (Guildf). 2016;100:102–110.
2018 with major in the fabrication and
[8] Rehman MM, Siddiqui GU, Gul JZ, et al. Resistive
characterization of 2D materials/poly-
switching in all-printed, flexible and hybrid MoS2-PVA
mers/nanocomposite based RRAMs,
nanocomposite based memristive device fabricated by
micro-sensing devices (humidity/tem-
reverse offset. Sci Rep. 2016;6:36195.
perature sensor). He is currently ser-
[9] Siddiqui GU, Rehman MM, Yang Y-J, et al. A
ving as an assistant professor in the Faculty of Electrical
two-dimensional hexagonal boron nitride/polymer
Engineering at the Ghulam Ishaq Khan Institute of
nanocomposite for flexible resistive switching
Engineering Sciences & Technology, Pakistan. His main
devices. J Mater Chem C. 2017;5:862–871.
research area is focused on fabrication and characterization
[10] Sarkar B, Lee B, Misra V. Understanding the gradual
of flexible electronic devices made of 2D materials, polymers
reset in Pt/Al2O3/Ni RRAM for synaptic applications.
and their nanocomposites.
Semicond Sci Technol. 2015;30:105014.
Mr. Hafiz Muhammad Mutee ur [11] Breuer T, Siemon A, Linn E, et al. A HfO2-based
Rehman received his MS degree in complementary switching crossbar adder. Adv
Mechatronics Engineering from Jeju Electron Mater. 2015;1:1–10.
National University, South Korea in [12] Rehman MM, Yang B-S, Yang Y-J, et al. Effect of device
2018. He is currently pursuing his structure on the resistive switching characteristics of
PhD degree in the Department of organic polymers fabricated through all printed
Electronics Engineering from the Jeju technology. Curr Appl Phys. 2017;17:533–540.
National University. His main research [13] Siddiqui GU, Rehman MM, Choi KH. Resistive
area is focused on the fabrication and switching phenomena induced by the heterostruc-
characterization of micro-electronic devices. ture composite of ZnSnO3 nanocubes interspersed
Sci. Technol. Adv. Mater. 21 (2020) 181 M.M. REHMAN et al.

ZnO nanowires. J Mater Chem C. [32] Jeong HY, Kim YI, Lee JY, et al. A low-temperature-
2017;5:5528–5537. grown TiO2-based device for the flexible stacked
[14] Hamaide G, Allibert F, Hovel H, et al. Impact of RRAM application. Nanotechnology. 2010;21:115203.
free-surface passivation on silicon on insulator [33] Xiao N, Villena MA, Yuan B, et al. Resistive random
buried interface properties by pseudotransistor access memory cells with a bilayer TiO2/SiO
characterization. J Appl Phys. 2007;101:114513. X insulating stack for simultaneous filamentary and
[15] Lee HY, Chen YS, Chen PS, et al. Evidence and solu- distributed resistive switching. Adv Funct Mater.
tion of over-RESET problem for HfOX based resistive 2017;27:1700384.
memory with sub-ns switching speed and high [34] Russo U, Kamalanathan D, Ielmini D, et al. Study of
endurance. Tech Dig - Int Electron Devices Meet multilevel programming in programmable metalliza-
IEDM; San Francisco, CA; 2010. p. 7–10. tion cell (PMC) memory. IEEE Trans Electron
[16] Ahn C, Jiang Z, Lee CS, et al. 1D selection device Devices. 2009;56:1040–1047.
using carbon nanotube FETs for high-density [35] Prócel LM, Trojman L, Moreno J, et al. Experimental
cross-point memory arrays. IEEE Trans Electron evidence of the quantum point contact theory in the
Devices. 2015;62:2197–2204. conduction mechanism of bipolar HfO2-based resis-
[17] Zhou J, Cai F, Wang Q, et al. Very low-programming- tive random access memories. J Appl Phys.
current RRAM with self-rectifying characteristics. 2013;114:074509.
IEEE Electron Device Lett. 2016;37:404–407. [36] Hong W-K, Jo G, Kwon -S-S, et al. Electrical proper-
[18] Xu N, Liu LF, Sun X, et al. Bipolar switching behavior ties of surface-tailored ZnO nanowire field-effect
in TiN/ZnO/Pt resistive nonvolatile memory with fast transistors. IEEE Trans Electron Devices.
switching and long retention. Semicond Sci Technol. 2008;55:3020–3029.
2008;23:075019. [37] Chen CY, Lin CA, Chen MJ, et al. ZnO/Al2O3 core–
[19] Kim GH, Lee JH, Ahn Y, et al. 32 × 32 crossbar array shell nanorod arrays: growth, structural characteriza-
resistive memory composed of a stacked Schottky tion, and luminescent properties. Nanotechnology.
diode and unipolar resistive memory. Adv Funct 2009;20:185605.
Mater. 2013;23:1440–1449. [38] Li QH, Gao T, Wang YG, et al. Adsorption and
[20] Ha SD, Ramanathan S. Adaptive oxide electronics: a desorption of oxygen probed from ZnO nanowire
review. J Appl Phys. 2011;110:071101. films by photocurrent measurements. Appl Phys
[21] Chang SH, Chae SC, Lee SB, et al. Effects of heat Lett. 2005;86:123117.
dissipation on unipolar resistance switching in Pt / [39] Wang X-F, Tian H, Zhao H-M, et al. Interface engi-
NiO/Pt capacitors. Appl Phys Lett. 2008;92:2008–2010. neering with MoS2-Pd nanoparticles hybrid structure
[22] Jana D, Samanta S, Maikap S, et al. Evolution of for a low voltage resistive switching memory. Small.
complementary resistive switching characteristics 2018;14:1702525.
using IrOx/GdOx/Al2O3/TiN structure. Appl Phys [40] Rehman MM, Siddiqui GU, Kim S, et al. Resistive
Lett. 2016;108:011605. switching effect in the planar structure of all-printed,
[23] Hanada A, Kinoshita K, Kishida S. Resistive switching flexible and rewritable memory device based on
by migration of hydrogen ions. Appl Phys Lett. advanced 2D nanocomposite of graphene quantum
2012;101:2010–2014. dots and white graphene flakes. J Phys D Appl Phys.
[24] Wei Z, Kanzawa Y, Arita K, et al. Highly reliable TaOx 2017;50:335104.
ReRAM and direct evidence of redox reaction [41] Qi YF, Shen ZJ, Zhao C, et al. Resistive switching beha-
mechanism. 2008 IEEE Int. Electron Devices Meet; vior of solution-processed AlOx and GO based RRAM
San Francisco, CA; 2008. p. 1–4. at low temperature. Solid State Electron. 2019;107735.
[25] Kim TW, Yang Y, Li F, et al. Electrical memory [42] Chen X, Zhu X, Zhang S-R, et al. Controlled nonvo-
devices based on inorganic/organic nanocomposites. latile transition in polyoxometalates-graphene oxide
NPG Asia Mater. 2012;4:e18. hybrid memristive devices. Adv Mater Technol.
[26] Tsunoda K, Fukuzumi Y, Jameson JR, et al. Bipolar 2019;4:1800551.
resistive switching in polycrystalline TiO2 films. Appl [43] Zhu K, Liang X, Yuan B, et al. Graphene–boron
Phys Lett. 2007;90:113501. nitride–graphene cross-point memristors with three
[27] Park K, Lee J-S. Reliable resistive switching memory stable resistive states. ACS Appl Mater Interfaces.
based on oxygen-vacancy-controlled bilayer structures. 2019;11:37999–38005.
RSC Adv. 2016;6:21736–21741. [44] Yalagala B, Khandelwal S,JD, Badhulika S. Wirelessly
[28] Takishita H, Ning S, Takeuchi K. 2015 silicon nanoe- destructible MgO-PVP-graphene composite based
lectronics workshop (SNW). 2015 Silicon flexible transient memristor for security applications.
Nanoelectron Work. 2015;2633:1–2. Mater Sci Semicond Process. 2019;104:104673.
[29] Moon HS, Kim JY, Jin HM, et al. Atomic layer deposi- [45] Chen T, Gao Y, Chen W, et al. Improved resistive
tion assisted pattern multiplication of block copoly- memory based on ZnO–graphene hybrids through
mer lithography for 5 nm scale nanopatterning. Adv redox process of graphene quantum dots. Phys
Funct Mater. 2014;24:4343–4348. Status Solidi – Rapid Res Lett. 2019;13:1900153.
[30] Huang JJ, Tseng YM, Luo WC, et al. One selector-one [46] Singh R, Kumar R, Kumar A, et al. Enhanced resistive
resistor (1S1R) crossbar array for high-density flexible switching in graphene oxide based composite thin
memory applications. Tech Dig - Int Electron Devices film for nonvolatile memory applications. Mater Res
Meet IEDM. 2011;733–736. Express. 2019;6:105621.
[31] Tsai C, Xiong F, Pop E, et al., Engineering C and [47] Choudhary S, Soni M, Sharma SK. Low voltage &
States U. Resistive random access memory enabled controlled switching of MoS2-GO resistive layers
by carbon nanotube. ACS Nano. 2013;7:5360– based ReRAM for non-volatile memory applications.
5366. Semicond Sci Technol. 2019;34:085009.
Sci. Technol. Adv. Mater. 21 (2020) 182 M.M. REHMAN et al.

[48] Prakash R, Sharma S, Kumar A, et al. Improved heterogeneous graphene-oxide interface. Adv Mater.
resistive switching performance in Cu-cation 2014;26:3275–3281.
migrated MoS2 based ReRAM device incorporated [66] Yang Y, Lee J, Lee S, et al. Oxide resistive memory
with tungsten nitride bottom electrode. Curr Appl with functionalized graphene as built-in selector
Phys. 2019;19:260–265. element. Adv Mater. 2014;26:3693–3699.
[49] Yang YJ, Rehman MM, Siddiqui GU, et al. Effect of [67] Jeon H, Park J, Jang W, et al. Detection of oxygen ion
adding a polymer and varying device size on the resistive drift in Pt/Al2O3/TiO2/Pt RRAM using interface-free
switching characteristics of perovskite nanocubes single-layer graphene electrodes. Carbon N Y.
heterojunction. Curr Appl Phys. 2017;17:1733–1741. 2014;75:209–216.
[50] Ali J, Rehman MM, Siddiqui GU, et al. Printing an [68] Zhao H, Tu H, Wei F, et al. Highly transparent
ITO-free flexible poly (4-vinylphenol) resistive switch- dysprosium oxide-based RRAM with multilayer gra-
ing device. Phys B Condens Matter. 2018;531:223–229. phene electrode for low-power nonvolatile memory
[51] Zhang F, Zhang H, Shrestha PR, et al. An ultra-fast application. IEEE Trans Electron Devices.
multi-level MoTe2-based RRAM. 2018 IEEE 2014;61:1388–1393.
International Electron Devices Meeting (IEDM) [69] Chakrabarti B, Roy T, Vogel EM. Nonlinear switching
(IEEE); Hilton San Francisco; 2018. p. 22.7.1–22.7.4. with ultralow reset power in graphene-insulator-
[52] Hou X, Pan R, Yu Q, et al. Tubular 3D resistive graphene forming-free resistive memories. IEEE
random access memory based on rolled-Up h- BN Electron Device Lett. 2014;35:750–752.
tube. Small. 2019;15:1803876. [70] Tian H, Zhao H, Wang X-F, et al. In situ tuning of
[53] Li Y, Sivan M, Niu JX, et al. Aerosol jet printed WSe2 switching window in a gate-controlled bilayer
based RRAM on kapton suitable for flexible mono- graphene-electrode resistive memory device. Adv
lithic memory integration. IEEE International Mater. 2015;27:7767–7774.
Conference on Flexible and Printable Sensors and [71] Lee S, Sohn J, Jiang Z, et al. Metal oxide-resistive
Systems (FLEPS) (IEEE); Glasgow; 2019. p. 1–3. memory using graphene-edge electrodes. Nat
[54] Das U, Bhattacharjee S, Mahato B, et al. Uniform, Commun. 2015;6:1–7.
large-scale growth of WS2 nanodomains via CVD [72] Jang BC, Seong H, Kim JY, et al. Ultra-low power,
technique for stable non-volatile RRAM application. highly uniform polymer memory by inserted multi-
Mater Sci Semicond Process. 2020;107:104837. layer graphene electrode. 2D Mater. 2015;2:044013.
[55] Sánchez-Salcedo S, Nieto A, Vallet-Regí M. [73] Jang CW, Hwang SW, Shin SH, et al. Significantly-
Hydroxyapatite/β-tricalcium phosphate/agarose macro- enhanced stabilities in flexible hybrid
porous scaffolds for bone tissue engineering. Chem Eng organic-inorganic perovskite resistive random access
J. 2008;137:62–71. memories by employing multilayer graphene trans-
[56] Sahu DP, Jammalamadaka SN. Detection of bovine parent conductive electrodes. J Korean Phys Soc.
serum albumin using hybrid TiO2+ graphene oxide 2018;73:934–939.
based bio-resistive random access memory device. Sci [74] Köymen I, Aydoğan Göktürk P, Kocabaş C, et al.
Rep. 2019;9:16141. Chemically addressed switching measurements in
[57] Liu S, Lu N, Zhao X, et al. Eliminating negative-SET graphene electrode memristive devices using in situ
behavior by suppressing nanofilament overgrowth in XPS. Faraday Discuss. 2019;213:231–244.
cation-based memory. Adv Mater. 2016;28:10623– [75] Dugu S, Pavunny SP, Limbu TB, et al. A graphene
10629. integrated highly transparent resistive switching
[58] Cheng CH, Chin A, Yeh FS. Novel ultra-low power memory device. APL Mater. 2018;6:058503.
RRAM with good endurance and retention. 2010 [76] Ott AK, Dou C, Sassi U, et al. Tetrahedral amorphous
Symposium on VLSI Technology (IEEE); Honolulu, carbon resistive memories with graphene-based
HI; 2010. p. 85–86. electrodes. 2D Mater. 2018;5:045028.
[59] Jiang J-W. Graphene versus MoS2: a short review. [77] Yao J, Lin J, Dai Y, et al. Highly transparent nonvo-
Front Phys. 2015;10:287–302. latile resistive memory devices from silicon oxide and
[60] Yu WJ, Chae SH, Lee SY, et al. Ultra-transparent, graphene. Nat Commun. 2012;3:1101.
flexible single-walled carbon nanotube non-volatile [78] Chakrabarti B, Roy T, Vogel EM. Nonlinear switching
memory device with an oxygen-decorated graphene with ultralow reset power in graphene-insulator-
electrode. Adv Mater. 2011;23:1889–1893. graphene forming-free resistive memories. IEEE
[61] Ji Y, Lee S, Cho B, et al. Flexible organic memory Electron Device Lett. 2014;35:750–752.
devices with multilayer graphene electrodes. ACS [79] Kapitanova OO, Panin GN, Kononenko OV, et al.
Nano. 2011;5:5995–6000. Resistive switching in graphene/graphene oxide/
[62] Park WI, Yoon JM, Park M, et al. Self-assembly- ZnO heterostructures. J Korean Phys Soc.
induced formation of high-density silicon oxide 2014;64:1399–1402.
memristor nanostructures on graphene and metal [80] Lu H, Lipatov A, Ryu S, et al. Ferroelectric tunnel
electrodes. Nano Lett. 2012;12:1235–1240. junctions with graphene electrodes. Nat Commun.
[63] Yao J, Lin J, Dai Y, et al. Highly transparent nonvo- 2014;5:5518.
latile resistive memory devices from silicon oxide and [81] Doh Y-J, Yi G-C. Nonvolatile memory devices based
graphene. Nat Commun. 2012;3:1101–1108. on few-layer graphene films. Nanotechnology.
[64] Lai Y-C, Hsu F-C, Chen J-Y, et al. Transferable and 2010;21:105204.
flexible label-like macromolecular memory on arbi- [82] Shin YJ, Kwon JH, Kalon G, et al. Ambipolar bistable
trary substrates with high performance and a facile switching effect of graphene. Appl Phys Lett.
methodology. Adv Mater. 2013;25:2733–2739. 2010;97:262105.
[65] Qian M, Pan Y, Liu F, et al. Tunable, ultralow-power [83] Son DI, Kim TW, Shim JH, et al. Flexible organic
switching in memristive devices enabled by a bistable devices based on graphene embedded in an
Sci. Technol. Adv. Mater. 21 (2020) 183 M.M. REHMAN et al.

insulating poly(methyl methacrylate) polymer layer. [102] El-Atab N, Cimen F, Alkis S, et al. Enhanced memory
Nano Lett. 2010;10:2441–2447. effect with embedded graphene nanoplatelets in ZnO
[84] Akdim B, Pachter R. Switching behavior of carbon charge trapping layer. Appl Phys Lett. 2014;105:2–6.
chains bridging graphene nanoribbons: effects of uni- [103] Cosmin Obreja A, Cristea D, Mihalache I, et al. Charge
axial strain. ACS Nano. 2011;5:1769–1774. transport and memristive properties of graphene quan-
[85] Wu C, Li F, Guo T, et al. Controlling memory effects tum dots embedded in poly(3-hexylthiophene) matrix.
of three-layer structured hybrid bistable devices Appl Phys Lett. 2014;105:083303.
based on graphene sheets sandwiched between two [104] Choi KH, Ali J, Doh YH. Exploring resistive switch-
laminated polymer layers. Org Electron. ing in poly(4-vinylphenol)–graphene nano-
2012;13:178–183. composite films. Jpn J Appl Phys. 2015;54:035103.
[86] Singh B, Mehta BR, Govind FX, et al. Electronic [105] Lee K, Hwang I, Lee S, et al. Enhancement of resistive
interaction and bipolar resistive switching in copper switching under confined current path distribution
oxide-multilayer graphene hybrid interface: graphene enabled by insertion of atomically thin defective
as an oxygen ion storage and blocking layer. Appl monolayer graphene. Sci Rep. 2015;5:11279.
Phys Lett. 2011;99:222109. [106] Transactions E C S and Society T E. Non-volatile
[87] Wu C, Li F, Zhang Y, et al. Recoverable electrical memory based on polymer-suspended graphene
transition in a single graphene sheet for application nanoplatelets with fractional and integer quantum
in nonvolatile memories. Appl Phys Lett. Conductance at 300K and zero magnetic field Y.
2012;100:042105. Kang. 2015;69:1–9.
[88] Ji Y, Choe M, Cho B, et al. Organic nonvolatile mem- [107] Ji Y, Lee S-A, Cha A-N, et al. Resistive switching
ory devices with charge trapping multilayer graphene characteristics of ZnO–graphene quantum dots and
film. Nanotechnology. 2012;23:105202. their use as an active component of an organic mem-
[89] He C, Shi Z, Zhang L, et al. Multilevel resistive switch- ory cell with one diode-one resistor architecture. Org
ing in planar graphene/SiO2 nanogap structures. ACS Electron. 2015;18:77–83.
Nano. 2012;6:4214–4221. [108] Kurkina II, Antonova IV, Nebogatikova NA, et al.
[90] Kim SM, Song EB, Lee S, et al. Transparent and Resistive switching effect and traps in partially fluorinated
flexible graphene charge-trap memory. ACS Nano. graphene films. J Phys D Appl Phys. 2016;49:095303.
2012;6:7879–7884. [109] Lee J, Du C, Sun K, et al. Tuning ionic transport in
[91] Lai Y-C, Wang D-Y, Huang I-S, et al. Low operation memristive devices by graphene with engineered
voltage macromolecular composite memory assisted nanopores. ACS Nano. 2016;10:3571–3579.
by graphene nanoflakes. J Mater Chem C. [110] Mannequin C, Delamoreanu A, Latu-Romain L, et al.
2013;1:552–559. Graphene-HfO2-based resistive RAM memories.
[92] Shindome A, Doioka Y, Beppu N, et al. Experimental Microelectron Eng. 2016;161:82–86.
study of two-terminal resistive random access memory [111] Ueda K, Aichi S, Asano H. Photo-controllable memris-
realized in mono- and multilayer exfoliated graphene tive behavior of graphene/diamond heterojunctions.
nanoribbons. Jpn J Appl Phys. 2013;52:04CN05. Appl Phys Lett. 2016;108:222102.
[93] Berzina TS, Gorshkov KV, Erokhin VV, et al. [112] Rani A, Kim DH. A mechanistic study on
Investigation of electrical properties of organic mem- graphene-based nonvolatile ReRAM devices. J Mater
ristors based on thin polyaniline-graphene films. Russ Chem C. 2016;4:11007–11031.
Microelectron. 2013;42:27–32. [113] Pan X, Skafidas E. Resonant tunneling based gra-
[94] Lee DU, Kim D, Kim EK, et al. Electrical character- phene quantum dot memristors. Nanoscale.
istics of resistive switching memory with metal oxide 2016;8:20074–20079.
nanoparticles on a graphene layer. Thin Solid Films. [114] Kapitanova OO, Panin GN, Cho HD, et al. Formation
2013;543:106–109. of self-assembled nanoscale graphene/graphene oxide
[95] He C, Li J, Wu X, et al. Tunable electroluminescence photomemristive heterojunctions using photocataly-
in planar graphene/SiO2 memristors. Adv Mater. tic oxidation. Nanotechnology. 2017;28:204005.
2013;25:5593–5598. [115] Shehata EM, Ibrahim MM, Balboul MR. Facile and
[96] Raghavan S, Stolichnov I, Setter N, et al. Long-term novel in-plane structured graphene/TiO2 nanocom-
retention in organic ferroelectric-graphene posites for memory applications. Adv Condens
memories. Appl Phys Lett. 2012;100:023507. Matter Phys. 2018;2018:1–9.
[97] Wang X, Xie W, Xu J-B. Graphene based non-volatile [116] Liu B, Liu Z, Chiu I-S, et al. Programmable synaptic
memory devices. Adv Mater. 2014;26:5496–5503. metaplasticity and below femtojoule spiking energy
[98] Rocci M, Tornos J, Rivera-Calzada A, et al. realized in graphene-based neuromorphic memristor.
Resistive switching in manganite/graphene hybrid ACS Appl Mater Interfaces. 2018;10:20237–20243.
planar nanostructures. Appl Phys Lett. [117] Sun Y, Wen D, Sun F. Eliminating negative-set beha-
2014;104:14–18. vior by adding a graphene blocking layer in resistive
[99] Li P, Jing G, Zhang B, et al. Single-crystalline mono- switching memory devices based on epoxy resin. Appl
layer and multilayer graphene nano switches. Appl Phys Express. 2019;12:074006.
Phys Lett. 2014;104:113110. [118] Tulsani SR, Rath AK, Late DJ. 2D-MoS2 nanosheets
[100] Khurana G, Misra P, Katiyar RS. Multilevel resistive as effective hole transport materials for colloidal PbS
memory switching in graphene sandwiched organic quantum dot solar cells. Nanoscale Adv.
polymer heterostructure. Carbon N Y. 2019;1:1387–1394.
2014;76:341–347. [119] Ji Y, Lee S-A, Cha A, et al. Resistive switching char-
[101] Wu C, Li F, Guo T. Efficient tristable resistive mem- acteristics of ZnO–graphene quantum dots and their
ory based on single layer graphene/insulating poly- use as an active component of an organic memory cell
mer multi-stacking layer. Appl Phys Lett. with one diode-one resistor architecture. Org Electron.
2014;104:183105. 2015;18:77–83.
Sci. Technol. Adv. Mater. 21 (2020) 184 M.M. REHMAN et al.

[120] Choi K-H, Ali J, Na K-H. Fabrication of [137] Wan T, Qu B, Du H, et al. Digital to analog resistive
graphene-nanoflake/poly(4-vinylphenol) polymer switching transition induced by graphene buffer layer
nanocomposite thin film by electrohydrodynamic in strontium titanate based devices. J Colloid
atomization and its application as flexible resistive Interface Sci. 2018;512:767–774.
switching device. Phys B Condens Matter. [138] Kim S, Jung HJ, Kim JC, et al. In situ observation of
2015;475:148–155. resistive switching in an asymmetric graphene oxide
[121] Vasu KS, Sampath S, Sood AK. Nonvolatile unipolar bilayer structure. ACS Nano. 2018;12:7335–7342.
resistive switching in ultrathin films of graphene and [139] Qi M, Bai L, Xu H, et al. Oxidized carbon quantum
carbon nanotubes. Solid State Commun. dot–graphene oxide nanocomposites for improving
2011;151:1084–1087. data retention of resistive switching memory.
[122] Cui P, Seo S, Lee J, et al. Nonvolatile memory device J Mater Chem C. 2018;6:2026–2033.
using gold nanoparticles covalently bound to reduced [140] Srivastava S, Dey P, Asapu S, et al. Role of GO and
graphene oxide. ACS Nano. 2011;5:6826–6833. r-GO in resistance switching behavior of bilayer TiO2
[123] Hu B, Quhe R, Chen C, et al. Electrically controlled based RRAM. Nanotechnology. 2018;29:505702.
electron transfer and resistance switching in reduced [141] Saini P, Singh M, Thakur J, et al. Probing the mechan-
graphene oxide noncovalently functionalized with ism for bipolar resistive switching in annealed gra-
thionine. J Mater Chem. 2012;22:16422. phene oxide thin films. ACS Appl Mater Interfaces.
[124] Liu J, Yin Z, Cao X, et al. Fabrication of flexible, 2018;10:6521–6530.
all-reduced graphene oxide non-volatile memory [142] Shi K, Wang Z, Xu H, et al. Complementary resistive
devices. Adv Mater. 2013;25:233–238. switching observed in graphene oxide-based memory
[125] Han S-T, Zhou Y, Yang QD, et al. Energy-band engi- device. IEEE Electron Device Lett. 2018;39:488–491.
neering for tunable memory characteristics through [143] Songkeaw P, Tunhoo B, Thiwawong T, et al.
controlled doping of reduced graphene oxide. ACS Transparent write-once-read-many-times memory
Nano. 2014;8:1923–1931. devices based on an ITO/EVA: rGO/ITOstructure.
[126] Ho NT, Senthilkumar V, Kim YS. Impedance spectro- J Mater Sci Mater Electron. 2018;29:17517–17524.
scopy analysis of the switching mechanism of reduced [144] Li L. Tunable memristic characteristics based on gra-
graphene oxide resistive switching memory. Solid phene oxide charge-trap memory. Micromachines.
State Electron. 2014;94:61–65. 2019;10:151.
[127] Kim H-D, Yun MJ, Lee JH, et al. Transparent [145] Zhao X, Ma J, Xiao X, et al. Breaking the
multi-level resistive switching phenomena observed current-retention dilemma in cation-based resistive
in ITO/RGO/ITO memory cells by the sol-gel switching devices utilizing graphene with controlled
dip-coating method. Sci Rep. 2015;4:4614. defects. Adv Mater. 2018;30:1705193.
[128] Myung S, Park J, Lee H, et al. Ambipolar memory [146] Liu G, Zhuang X, Chen Y, et al. Bistable electrical
devices based on reduced graphene oxide and switching and electronic memory effect in a
nanoparticles. Adv Mater. 2010;22:2045–2049. solution-processable graphene oxide-donor polymer
[129] Zhuang X-D, Chen Y, Liu G, et al. Conjugated- complex. Appl Phys Lett. 2009;95:253301.
polymer-functionalized graphene oxide: synthesis [147] Hong SK, Kim J-E, Kim SO, et al. Non-volatile mem-
and nonvolatile rewritable memory effect. Adv ory using graphene oxide for flexible electronics. 10th
Mater. 2010;22:1731–1735. IEEE International Conference on Nanotechnology
[130] Zhang B, Liu G, Chen Y, et al. Conjugated (IEEE); Seoul, Korea; 2010. p. 604–606.
polymer-grafted reduced graphene oxide for nonvo- [148] Jeong HY, Kim JY, Kim JW, et al. Graphene oxide
latile rewritable memory. Chem - A Eur J. thin films for flexible nonvolatile memory
2011;17:13646–13652. applications. Nano Lett. 2010;10:4381–4386.
[131] Rani A, Song J-M, Jung Lee M, et al. Reduced gra- [149] Wang L-H, Yang W, Sun Q-Q, et al. The mechanism
phene oxide based flexible organic charge trap mem- of the asymmetric SET and RESET speed of graphene
ory devices. Appl Phys Lett. 2012;101:233308. oxide based flexible resistive switching memories.
[132] Lin CL, Chang WY, Huang YL, et al. Resistance Appl Phys Lett. 2012;100:063509.
switching behavior of ZnO resistive random access [150] Kim I, Siddik M, Shin J, et al. Low temperature
memory with a reduced graphene oxide capping solution-processed graphene oxide/Pr 0.7 Ca 0.3
layer. Jpn J Appl Phys. 2015;54:04DJ08. MnO3 based resistive-memory device. Appl Phys
[133] Pradhan SK, Xiao B, Mishra S, et al. Resistive switch- Lett. 2011;99:042101.
ing behavior of reduced graphene oxide memory cells [151] Wu C, Li F, Zhang Y, et al. Highly reproducible
for low power nonvolatile device application. Sci Rep. memory effect of organic multilevel resistive-switch
2016;6:26763. device utilizing graphene oxide sheets/polyimide
[134] Cao Y, Fu Y, Li D, et al. Organophosphorus-based hybrid nanocomposite. Appl Phys Lett.
polymer covalently functionalized reduced graphene 2011;99:042108.
oxide: in situ synthesis and nonvolatile memory [152] Zhang B, Liu G, Chen Y, et al. Conjugated
effect. Carbon N Y. 2019;141:758–767. polymer-grafted reduced graphene oxide for nonvo-
[135] Zhang B, Liu L, Wang L, et al. Covalent modification of latile rewritable memory. Chem - A Eur J.
graphene oxide with poly(N-vinylcarbazole) containing 2011;17:13646–13652.
pendant azobenzene chromophores for nonvolatile [153] Yu A-D, Liu C-L, Chen W-C. Supramolecular block
ternary memories. Carbon N Y. 2018;134:500–506. copolymers: graphene oxide composites for memory
[136] Choi J-Y, Yu H-C, Lee J, et al. Preparation of poly- device applications. Chem Commun. 2012;48:383–385.
imide/graphene oxide nanocomposite and its applica- [154] Wang Z, Tjoa V, Wu L, et al. Mechanism of different
tion to nonvolatile resistive memory device. Polymers switching directions in graphene oxide based RRAM.
(Basel). 2018;10:901. J Electrochem Soc. 2012;159:K177–82.
Sci. Technol. Adv. Mater. 21 (2020) 185 M.M. REHMAN et al.

[155] Mustafa M, Awais MN, Pooniah G, et al. Electrospray [172] Kim SK, Kim JY, Choi S-Y, et al. Direct observation of
deposition of a graphene-oxide thin film, its charac- conducting nanofilaments in graphene-oxide-
terization and investigation of its resistive switching resistive switching memory. Adv Funct Mater.
performance. J Korean Phys Soc. 2012;61:470–475. 2015;25:6710–6715.
[156] Venugopal G, Kim S-J. Observation of nonvolatile [173] Aleshin AN, Krylov PS, Berestennikov AS, et al. The
resistive memory switching characteristics in Ag/ redox nature of the resistive switching in nanocom-
Graphene-oxide/Ag devices. J Nanosci Nanotechnol. posite thin films based on graphene (graphene oxide)
2012;12:8522–8525. nanoparticles and poly(9-vinylcarbazole). Synth Met.
[157] Bhunia P, Hwang E, Min M, et al. A non-volatile 2016;217:7–13.
memory device consisting of graphene oxide cova- [174] Nagareddy VK, Barnes MD, Zipoli F, et al. Multilevel
lently functionalized with ionic liquid. Chem ultrafast flexible nanoscale nonvolatile hybrid gra-
Commun. 2012;48:913–915. phene oxide–titanium oxide memories. ACS Nano.
[158] Seo S, Yoon Y, Lee J, et al. Nitrogen-doped partially 2017;11:3010–3021.
reduced graphene oxide rewritable nonvolatile [175] Valanarasu S, Kulandaisamy I, Kathalingam A, et al.
memory. ACS Nano. 2013;7:3607–3615. High-performance memory device using graphene
[159] Zhang B, Chen Y, Ren Y, et al. in situ synthesis and oxide flakes sandwiched polymethylmethacrylate
nonvolatile rewritable-memory effect of polyaniline- layers. J Nanosci Nanotechnol. 2013;13:6755–6759.
functionalized graphene oxide. Chem - A Eur J. [176] Liu J, Zeng Z, Cao X, et al. Preparation of MoS2-
2013;19:6265–6273. polyvinylpyrrolidone nanocomposites for flexible non-
[160] Senthilkumar V, Kathalingam A, Valanarasu S, et al. volatile rewritable memory devices with reduced gra-
Bipolar resistive switching of solution processed TiO2– phene oxide electrodes. Small. 2012;8:3517–3522.
graphene oxide nanocomposite for nonvolatile memory [177] Xu X-Y, Yin Z-Y, Xu C-X, et al. Resistive switching
applications. Phys Lett A. 2013;377:2432–2435. memories in MoS2 nanosphere assemblies. Appl Phys
[161] Yi M, Cao Y, Ling H, et al. Temperature dependence Lett. 2014;104:033504.
of resistive switching behaviors in resistive random [178] Sun B, Zhao WX, Liu YH, et al. Resistive switching
access memory based on graphene oxide film. effect of Ag/MoS2/FTO device. Funct Mater Lett.
Nanotechnol. 2014;25:185202. 2015;8:4.
[162] Yun DY, Park HM, Kim SW, et al. Enhancement of [179] Bessonov AA, Kirikova MN, Petukhov DI, et al.
memory margins for stable organic bistable devices Layered memristive and memcapacitive switches for
based on graphene-oxide layers due to embedded printable electronics. Nat Mater. 2015;14:199–204.
CuInS2 quantum dots. Carbon N Y. 2014;75:244–248. [180] Sangwan VK, Jariwala D, Kim IS, et al. Gate-tunable
[163] Khurana G, Misra P, Kumar N, et al. Tunable power memristive phenomena mediated by grain bound-
switching in nonvolatile flexible memory devices aries in single-layer MoS2. Nat Nanotechnol.
based on graphene oxide embedded with ZnO 2015;10:403–406.
nanorods. J Phys Chem C. 2014;118:21357–21364. [181] Cheng P, Sun K, Hu YH. Memristive behavior and
[164] Chung JB, Bae YC, Lee AR, et al. Enhancement of ideal memristor of 1T phase MoS2 nanosheets. Nano
memory windows in Pt/Ta2O5−x/Ta bipolar resistive Lett. 2016;16:572–576.
switches via a graphene oxide insertion layer. Thin [182] Zhang P, Gao C, Xu B, et al. Structural phase transi-
Solid Films. 2015;587:57–60. tion effect on resistive switching behavior of MoS2
[165] Midya A, Gogurla N, Ray SK. Flexible and transpar- -polyvinylpyrrolidone nanocomposites films for flex-
ent resistive switching devices using Au nanoparticles ible memory devices. Small. 2016;12:2077–2084.
decorated reduced graphene oxide in polyvinyl alco- [183] Wang W, Panin GN, Fu X, et al. MoS2 memristor
hol matrix. Curr Appl Phys. 2015;15:706–710. with photoresistive switching. Sci Rep. 2016;6:31224.
[166] Yun DY, Kim TW. Nonvolatile memory devices [184] Fan F, Zhang B, Cao Y, et al. Solution-processable
based on Au/graphene oxide nanocomposites with poly(N-vinylcarbazole)-covalently grafted MoS2
bilateral multilevel characteristics. Carbon N Y. nanosheets for nonvolatile rewritable memory
2015;88:26–32. devices. Nanoscale. 2017;9:2449–2456.
[167] Harris P, Salimian A, Banerjee I, et al. Graphene oxide [185] Wang D, Ji F, Chen X, et al. Quantum conductance in
thin films for resistive memory switches. IET Circuits MoS2 quantum dots-based nonvolatile resistive mem-
Devices Syst. 2015;9:428–433. ory device. Appl Phys Lett. 2017;110:093501.
[168] Kim YN, Lee NH, Yun DY, et al. Multilevel charac- [186] Zhou GD, Lu ZS, Yao YQ, et al. Mechanism for
teristics and operating mechanisms of nonvolatile bipolar resistive switching memory behaviors of a
memory devices based on a floating gate of graphene self-assembled three-dimensional MoS2 microsphere
oxide sheets sandwiched between two polystyrene composed active layer. J Appl Phys. 2017;121:155302.
layers. Org Electron. 2015;25:165–169. [187] Li G, Tan F, Lv B, et al. Thiol-modified MoS2
[169] Thakre A, Borkar H, Singh BP, et al. Electroforming free nanosheets as a functional layer for electrical bistable
high resistance resistive switching of graphene oxide devices. Opt Commun. 2018;406:112–117.
modified polar-PVDF. RSC Adv. 2015;5:57406–57413. [188] Xia Y, Sun B, Wang H, et al. Metal ion formed con-
[170] Wu H-Y, Lin -C-C, Lin C-H. Characteristics of ductive filaments by redox process induced nonvola-
graphene-oxide-based flexible and transparent resis- tile resistive switching memories in MoS2 film. Appl
tive switching memory. Ceram Int. 2015;41:S823–8. Surf Sci. 2017;426:812–816.
[171] Khurana G, Misra P, Kumar N, et al. Enhanced resis- [189] Bhattacharjee S, Sarkar PK, Prajapat M, et al. Electrical
tive switching in forming-free graphene oxide films reliability, multilevel data storage and mechanical stabi-
embedded with gold nanoparticles deposited by lity of MoS2-PMMA nanocomposite-based non-volatile
electrophoresis. Nanotechnology. 2016;27:015702. memory device. J Phys D Appl Phys. 2017;50:265103.
Sci. Technol. Adv. Mater. 21 (2020) 186 M.M. REHMAN et al.

[190] Li D, Wu B, Zhu X, et al. MoS2 memristors exhibiting based on MoSe2-doped ultralong Se microwires.
variable switching characteristics toward biorealistic Appl Phys Lett. 2016;109:143904.
synaptic emulation. ACS Nano. 2018;12:9240–9252. [207] Han P, Sun B, Li J, et al. Ag filament induced non-
[191] Chen P-A, Ge R-J, Lee J-W, et al. An RRAM with volatile resistive switching memory behaviour in hex-
a 2D material embedded double switching layer for agonal MoSe2 nanosheets. J Colloid Interface Sci.
neuromorphic computing. 2018 IEEE 13th 2017;505:148–153.
Nanotechnology Materials and Devices Conference [208] Li P, Sun B, Zhang X, et al. Effect of temperature on
(NMDC) vol 1 (IEEE); Portland, OR. 2018; p. 1–4. the magnetism and memristive memory behavior of
[192] Hou X, Yan X, Liu C, et al. Operation mode switch- MoSe2 nanosheets. Mater Lett. 2017;202:13–16.
able charge-trap memory based on few-layer MoS2. [209] Xia Y, Guo XG, Ji TX, et al. Neutrophil count to
Semicond Sci Technol. 2018;33:034001. lymphocyte count ratio is a potential diagnostic
[193] Kumar A, Pawar S, Sharma S, et al. Bipolar resistive index for bacteremia in adult. Life Sci J.
switching behavior in MoS2 nanosheets fabricated on 2014;11:172–177.
ferromagnetic shape memory alloy. Appl Phys Lett. [210] Qian K, Tay RY, Nguyen VC, et al. Hexagonal boron
2018;112:262106. nitride thin film for flexible resistive memory
[194] Kim M, Ge R, Wu X, et al. Zero-static power applications. Adv Funct Mater. 2016;26:2176–2184.
radio-frequency switches based on MoS2 [211] Jing X, Puglisi F, Akinwande D, et al. Chemical vapor
atomristors. Nat Commun. 2018;9:2524. deposition of hexagonal boron nitride on
[195] Qiu JT, Samanta S, Dutta M, et al. Controlling resis- metal-coated wafers and transfer-free fabrication of
tive switching by using an optimized MoS2 interfacial resistive switching devices. 2D Mater. 2019;6:035021.
layer and the role of top electrodes on ascorbic acid [212] Das U, Mahato B, Sarkar PK, et al. Bipolar resistive
sensing in TaO x-based RRAM. Langmuir. switching behaviour of WS2 thin films grown by
2019;35:3897–3906. chemical vapour deposition. Dae Solid State Physics
[196] Das U, Bhattacharjee S, Sarkar PK, et al. A multi-level Symposium 2018 vol 2115; Hisar, Haryana; 2019. p.
bipolar memristive device based on visible light sensing 030274.
MoS2 thin film mater. Res Express. 2019;6:075037. [213] Rehman MM, Siddiqui GU, Doh YH, et al. Highly
[197] Feng X, Li Y, Wang L, et al. First demonstration of a flexible and electroforming free resistive switching
fully-printed MoS2 RRAM on flexible substrate with behavior of tungsten disulfide flakes fabricated
ultra-low switching voltage and its application as through advanced printing technology. Semicond
electronic synapse. 2019 Symposium on VLSI Sci Technol. 2017;32:095001.
Technology (IEEE); Kyoto, Japan; 2019. p. T88–9. [214] Yin Z, Zeng Z, Liu J, et al. Memory devices using
[198] Zhai Y, Yang X, Wang F, et al. Infrared-sensitive a mixture of MoS2 and graphene oxide as the active
memory based on direct-grown MoS2-upconversion- layer. Small. 2013;9:727–731.
nanoparticle heterostructure. Adv Mater. 2018;30: [215] Shinde SM, Kalita G, Tanemura M. Fabrication of
1803563. poly(methyl methacrylate)-MoS2 /graphene hetero-
[199] Sun B, Zhao W, Liu Y, et al. Resistive switching effect of structure for memory device application. J Appl
Ag/MoS2/FTO device. Funct Mater Lett. 2015;8:2–5. Phys. 2014;116:214306.
[200] Rehman MM, Siddiqui GU, Gul JZ, et al. Resistive [216] Shin GH, Kim C-K, Bang GS, et al. Multilevel resistive
switching in all-printed, flexible and hybrid MoS2 switching nonvolatile memory based on MoS2
-PVA nanocomposite based memristive device fabri- nanosheet-embedded graphene oxide. 2D Mater.
cated by reverse. Offset Sci Rep. 2016;6:36195. 2016;3:034002.
[201] Kadhim MS, Yang F, Sun B, et al. Existence of resis- [217] Wu L, Guo J, Zhong W, et al. Flexible, multilevel, and
tive switching memory and negative differential resis- low-operating-voltage resistive memory based on
tance state in self-colored MoS2/ZnO heterojunction MoS2–rGO hybrid. Appl Surf Sci. 2019;463:947–952.
devices. ACS Appl Electron Mater. 2019;1:318–324. [218] Yalagala B, Sahatiya P, Mattela V, et al. Ultra-low cost,
[202] Wang Q, Wen Y, Cai K, et al. Nonvolatile infrared large area graphene/MoS2-based piezotronic memris-
memory in MoS2/PbS van der waals heterostructures. tor on paper: a systematic study for both direct cur-
Sci Adv. 2018;4:eaap7916. rent and alternating current inputs. ACS Appl
[203] Yan Y, Sun B, Ma D. Resistive switching memory Electron Mater. 2019;1:883–891.
characteristics of single MoSe2 nanorods. Chem [219] Li Y, Long S, Liu Q, et al. Resistive switching perfor-
Phys Lett. 2015;638:103–107. mance improvement via modulating nanoscale con-
[204] Han P, Sun B, Cheng S, et al. Preparation of MoSe2 ductive filament, involving the application of
nano-islands array embedded in a TiO2 matrix for two-dimensional layered materials. Small.
photo-regulated resistive switching memory. J Alloys 2017;13:1604306.
Compd. 2016;664:619–625. [220] Hong SK, Kim JE, Kim SO, et al. Flexible resistive
[205] Zhang X, Qiao H, Nian X, et al. Resistive switching switching memory device based on graphene oxide.
memory behaviours of MoSe2 nano-islands array. IEEE Electron Device Lett. 2010;31:1005–1007.
J Mater Sci Mater Electron. 2016;27:7609–7613. [221] Lin -C-C, Wu H-Y, Lin N-C, et al. Graphene-oxide-
[206] Zhou G, Sun B, Yao Y, et al. Investigation of the based resistive switching device for flexible nonvolatile
behaviour of electronic resistive switching memory memory application. Jpn J Appl Phys. 2014;53:05FD03.