Memristive Neuromorphic Interfaces: Integrating Sensory Modalities with Artificial Neural Networks

Abstract

The advent of the Internet of Things (IoT) has led to an exponential growth in data generated from sensors. Consequently, a time- and energy-efficient method for processing complex and unstructured...

Full text

rsc.li/materials-horizons
Materials
Horizons
rsc.li/materials-horizons
ISSN 2051-6347
COMMUNICATION
Blaise L. Tardy, Orlando J. Rojas et al.
Biofabrication of multifunctional nanocellulosic 3D structures:
a facile and customizable route
Volume 5
Number 3
May 2018
Pages 311-580
Materials
Horizons
This is an Accepted Manuscript, which has been through the
Royal Society of Chemistry peer review process and has been
accepted for publication.
Accepted Manuscripts are published online shortly after acceptance,
before technical editing, formatting and proof reading. Using this free
service, authors can make their results available to the community, in
citable form, before we publish the edited article. We will replace this
Accepted Manuscript with the edited and formatted Advance Article as
soon as it is available.
You can find more information about Accepted Manuscripts in the
Information for Authors.
Please note that technical editing may introduce minor changes to the
text and/or graphics, which may alter content. The journal’s standard
Terms & Conditions and the Ethical guidelines still apply. In no event
shall the Royal Society of Chemistry be held responsible for any errors
or omissions in this Accepted Manuscript or any consequences arising
from the use of any information it contains.
Accepted Manuscript
View Article Online
View Journal
This article can be cited before page numbers have been issued, to do this please use: J. E. Kim, K. Soh,
S. Hwang, D. Y. Yang and J. H. Yoon, Mater. Horiz., 2025, DOI: 10.1039/D5MH00038F.

Wider impact
The implementation of artificial sensory systems is essential for converting vast amounts of
environmental information into input signals required for neuromorphic computing. When
realized using memristors, such systems effectively compress signals during the conversion
process while retaining adaptive, nociceptive, and spatiotemporal information critical for
learning and inference. Furthermore, their compatibility with a wide range of sensors ensures
excellent expandability, while the dynamic resistive switching properties of memristors enable
diverse signal conversion strategies. Memristor-based artificial sensory systems not only
emulate human sensory processing but also offer significant advantages in terms of energy
efficiency and miniaturization, making them highly suitable for edge computing and wearable
technologies. Their ability to perform parallel signal processing can also enhance real-time
decision-making in complex environments. Gaining insights into memristor-based artificial
sensory systems, which process patterned sensory data akin to human perception, can drive
future advancements in neuromorphic computing, industrial automation, and robotics.
Page 1 of 30 Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Journal Name ARTICLE
This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00 , 1 -3 | 1
Please do not adjust margins
Please do not adjust margins
Data availability statement
No primary research results, software or code have been included and no new data were generated or analysed as part of this
review.
Page 2 of 30Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

REVIEW
Please do not adjust margins
Please do not adjust margins
Received 00th January 20xx,
Accepted 00th January 20xx
DOI: 10.1039/x0xx00000x
Memristive Neuromorphic Interfaces: Integrating Sensory
Modalities with Artificial Neural Networks
Ji Eun Kima,b†, Keunho Sohc†, Suin Hwangc, Do Young Yangc, and Jung Ho Yoonc*
The advent of the Internet of Things (IoT) has led to exponential growth in data generated from sensors, requiring efficient
methods to process complex and unstructured external information. Unlike conventional von Neumann sensory systems
with separate data collection and processing units, biological sensory systems integrate sensing, memory, and computing to
process environmental information in real time with high efficiency. Memristive neuromorphic sensory systems using
memristors as their basic components have emerged as promising alternatives to CMOS-based systems. Memristors can
closely replicate the key characteristics of biological receptors, neurons, and synapses by integrating the threshold and
adaptation properties of receptors, the action potential firing in neurons, and the synaptic plasticity of synapses.
Furthermore, through careful engineering of their switching dynamics, the electrical properties of memristors can be
tailored to emulate specific functions, while benefiting from high operational speed, low power consumption, and
exceptional scalability. Consequently, their integration with high-performance sensors offers a promising pathway toward
realizing fully integrated artificial sensory systems that can efficiently process and respond to diverse environmental stimuli
in real time. In this review, we first introduce the fundamental principles of memristive neuromorphic technologies for
artificial sensory systems, explaining how each component is structured and what functions they perform. We then discuss
how these principles can be applied to replicate the four traditional senses, highlighting the underlying mechanisms and
recent advances in mimicking biological sensory functions. Finally, we address the remaining challenges and provide
prospects for the continued development of memristor-based artificial sensory systems.
11. Introduction
2The growing demand for automation in supply chains,
3manufacturing, robotics, and unmanned vehicles has driven the
4development of artificial intelligence (AI) technologies. These
5technologies have the potential to significantly improve efficiency
6and autonomy across various industries using sensory systems
7comprising sensors and computational networks to sense the
8surroundings and acquire information from the environment in real
9time.1, 2 For instance, conventional complementary metal-oxide
10 semiconductor (CMOS)-based systems have demonstrated
11 intelligent recognition and control applications, such as image
12 classification, natural language processing, and decision-making
13 tasks.3-10 However, because the von Neumann architecture
14 physically separates memory and processing units, conventional
15 systems require massive amounts of data transfer between them.
16 This results in high power consumption and causes significant
17 latency, commonly referred to as the von Neumann bottleneck,
18 which fundamentally degrades the performance of AI applications.11-
19 14
20 Unlike conventional systems, biological sensory systems detect,
21 interpret, and store external information in a data-parallel and
22 integrated manner.15 This is enabled by receptors that generate
23 electrical signals only when stimuli exceed a threshold, selectively
24 adapting to harmless, repetitive inputs. These signals are transmitted
25 as action potentials (spikes) through neurons to specific brain regions,
26 where they are processed in an event-driven, adaptive, and parallel
27 manner, enabling learning and inference.16, 17 Inspired by the energy-
28 efficient and fault-tolerant nature of biological systems,
29 neuromorphic computing has been developed to overcome the
30 technical limitations of conventional CMOS-based systems.18-21 It
31 supports the integration, processing, and storage of sensory
32 information, playing a crucial role in advanced functions, such as
33 decision-making, cognition, learning, and memory. Moreover,
34 neuromorphic computing can execute multiple tasks simultaneously
35 in highly parallel settings with a low power consumption of 1–100 fJ
36 per synaptic event.22 The exceptional capabilities of memristors
37 enable their integration with neuromorphic learning algorithms to
38 facilitate advanced functions. Large-scale integration and hardware
39 implementation using CMOS-compatible processes are essential to
40 leverage these capabilities, with extensive research currently
41 underway. The technology has now advanced beyond hybrid 1T1R
42 structures, reaching a stage where fully memristor-based hardware
43 implementations are feasible. This progress has demonstrated the
44 practical applicability of memristors across various AI applications,
45 validating their potential for widespread deployment.23-27
46 Therefore, it is crucial to implement artificial sensory systems
47 capable of mimicking the roles of biological receptors, neurons, and
a.Electronic Materials Research Center, Korea Institute of Science and Technology
(KIST), Seoul 02791 Republic of Korea Address here.
b.Department of Materials Science and Engineering, Korea University, Seoul 02841,
Republic of Korea
c. School of Advanced Materials and Engineering, Sungkyunkwan University (SKKU),
Suwon 16419, Republic of Korea
* E-mail: junghoyoon@skku.edu
† Contributed equally to this work
Page 3 of 30 Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Review Journal Name
2 | J. Name., 2012, 00 , 1- 3 This journal is © The Royal Society of Chemistry 20xx
Please do not adjust margins
Please do not adjust margins
1synapses to fully leverage neuromorphic computing.28-31 Although
2conventional CMOS-based electronics have been used to develop
3artificial synapses and neurons as neuromorphic devices, they are
4limited by circuit area and energy efficiency.32-34 Since the CMOS-
5based devices are optimized for digital switching, they struggle to
6handle smooth and continuous signal variations, which are essential
7for accurately reflecting external stimuli. Thus, essential functions
8such as the accumulation of external stimuli, the generation of
9corresponding output signals, and information storage inevitably be
10 performed by separate components. As a result, the emulation
11 process compromises both area and energy efficiency in proportion
12 to the number of devices used.35 Moreover, implementing analog
13 switching to achieve both the precision and dynamic range required
14 for emulating biological counterparts remains a significant challenge
15 in conventional CMOS-based systems. These systems necessitate the
16 incorporation of additional circuitry, such as Digital-to-Analog
17 Converters (DACs), to facilitate analog switching. Although more
18 complex DAC configurations are required to enhance the output
19 resolution, the resulting output often lacks the desired smoothness.
20 Meanwhile, among various neuromorphic devices, the memristor
21 stands out for its area-efficient structure as well as high-speed and
22 low-power operation. Additionally, their excellent scalability,
23 durability, and uniformity make them well-suited for the reliable
24 implementation of artificial sensory systems.36-40 Furthermore, a
25 unique attribute of memristors is their ability to gradually switch
26 between a low-resistance state (LRS) and a high-resistance state
27 (HRS) in response to external stimuli, such as voltage or current. In
28 other words, memristors exhibit continuous and dynamic resistive
29 state changes rather than relying on binary resistance states. This
30 enables the direct processing of analog external stimuli without the
31 complex configuration of using multiple devices or peripheral circuits
32 such as analog-to-digital converters. Therefore, the dynamic resistive
33 switching provided by memristors is essential for replicating the
34 artificial sensory system, as it more efficiently captures the full
35 fidelity of incoming signals. Owing to these advantages, memristors
36 have been widely utilized in the implementation of artificial
37 receptors, synapses, and neurons.41, 42 In particular, their material
38 composition, device structure, and switching dynamics can be
39 carefully engineered to optimize switching behavior, making them
40 adaptable to both volatile and non-volatile properties—key
41 characteristics for mimicking biological elements.34, 43-51 Thus,
42 integrating memristive devices with various sensors facilitates the
43 implementation of artificial sensory systems corresponding to
44 tactile, visual, auditory, and olfactory modalities.52, 53
45 In biological sensory systems, sensory receptors located in the
46 sensory organs convert external perceptual signals into receptor
47 potentials, and sensory neurons integrate these potentials to initiate
48 action potentials. Finally, the synapses store the encoded sensory
49 information. Similarly, in a bioinspired memristive sensory system,
50 sensors generally convert external stimuli into electrical signals,
51 which are then applied to memristors. Subsequently, the memristive
52 receptor device that receives the signal generates a potential that is
53 proportional to the input, incorporates information regarding
54 harmful stimuli, and transfers it to the subsequent sensory system.
55 Subsequently, the integrated memristive synapse and neural devices
56 respond to input signals in a manner analogous to biological
57 perception systems. By mimicking the biological sensory system, the
58 integration of sensory, processing, and memory components in
59 bioinspired memristive systems enables high power efficiency, low
60 latency, and excellent processing capabilities.
61 Despite the versatility of memristors, current research has
62 predominantly focused on signal conversion based on their switching
63 characteristics. This approach has contributed immensely to the
64 advancement of neuromorphic computing by enabling reliable and
65 direct conversion of external stimuli into signals that drive neural
66 networks implemented in hardware and software. However, studies
67 on how closely these conversions align with the behavior of the
68 human nervous system are lacking. The existing memristor-based
69 systems often fail to fully capture the intricate dynamics of biological
70 sensory systems, particularly in terms of complexity and adaptability.
71 Devices capable of replicating the full range of functions of biological
72 receptors, neurons, and synapses remain exceedingly rare. Even at
73 the individual level, most artificial systems struggle to replicate all
74 the critical functions of a single biological element. In artificial
75 sensory systems, this limitation is further compounded by the
76 frequent exclusion of specific functions or entire elements, resulting
77 in incomplete or inefficient performance. This highlights a critical
78 challenge: implementing all essential characteristics necessary for
79 effective emulation. For artificial sensory systems to accurately
80 process external stimuli across diverse environmental conditions,
81 several crucial properties must be considered, including sensitivity,
82 adaptability, and spatiotemporal processability. For instance,
83 biological systems can dynamically adjust their sensitivity to external
84 stimuli, such as by enhancing auditory perception in noisy
85 environments or modulating visual processing under low light.
86 Emulating this adaptability requires devices capable of self-tuning
87 and learning in response to changing environmental conditions.
88 Moreover, processing spatiotemporal patterns—similar to biological
89 synapses responding to time-dependent signals—remains essential
90 for replicating complex sensory functions. A systematic
91 understanding of these properties is fundamental to developing
92 artificial sensory systems that process complex input patterns with
93 greater accuracy and efficiency.
94 In this review, the recent advances, challenges, and prospects of bio-
95 inspired memristive artificial sensory systems are comprehensively
96 examined. In this context, the switching performance metrics
97 required for memristors in the implementation of artificial sensory
98 systems, as depicted in Fig. 1, along with the sensory modalities they
99 aim to emulate, are discussed. The subsequent sections first explore
100 the fundamental roles of receptors, neurons, and synapses in
101 biological sensory systems, along with the corresponding switching
102 characteristics of memristors essential for replicating these neuronal
103 components. Next, innovative cases of bio-inspired artificial sensory
104 systems developed for the four primary senses—tactile, visual,
105 auditory, and olfactory—are presented. Recent memristor research
106 progress is then examined, focusing on how closely these systems
107 mimic biological sensory functions and evaluating the effectiveness
108 of these advancements. Finally, challenges and prospects for the
109 development of memristor-based artificial sensory systems are
110 addressed. This review aims to encourage ongoing research and
Page 4 of 30Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Journal Name Review
This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00 , 1 -3 | 3
Please do not adjust margins
Please do not adjust margins
1development, fostering a deeper understanding and broader
2application of bio-inspired sensory systems by analyzing the roles of
3receptors, neurons, and synapses, the switching dynamics of
4memristors, and the necessary characteristics for each type of neural
5implementation.
62. Element of the nervous system: Receptor,
7Neuron, and Synapse
8To emulate the characteristics of receptors, neurons, and synapses
9using memristors, a comprehensive understanding of their
10 operational mechanisms is required. Additionally, investigating the
11 switching properties of memristors and exploring how these
12 properties can be utilized to mimic each component are essential.
13 This process is crucial for precisely controlling the electrical
14 characteristics of memristors and effectively reproducing the
15 complex functions of the nervous system, as shown in Fig. 2.
16 2.1 Receptor
17 Receptors play a crucial role in detecting and responding to various
18 stimuli, enabling us to perceive and interact with the environment.54,
19 55 Receptors convert physical and chemical stimuli into electrical
20 signals. This process enables humans to appropriately respond to
21 stimuli. Receptors have evolved to be specifically responsive to
22 stimuli and can be classified into categories based on their ability to
23 accommodate different external stimuli, such as mechanoreceptors,
24 thermoreceptors, photoreceptors, chemoreceptors, and
25 nociceptors.
26 Receptors operate based on thresholds and relaxation.56 The
27 threshold indicates the minimum intensity of a stimulus required to
28 be activated, below which the receptor remains unresponsive. This
29 characteristic enables the receptors to filter out insignificant minor
30 stimuli and focus on more critical signals. Upon activation by external
31 stimuli, receptors transition into a relaxed state where their
32 responsiveness to the stimulus gradually diminishes, enabling them
33 to revert to their initial state. During the relaxation state, receptors
34 retain a certain degree of activation; consequently, the threshold
35 intensity of the stimulus for reactivation is reduced compared with
36 that of the initial activation. This phenomenon, known as
37 sensitization, is crucial for modulating receptor sensitivity.57
38 Additionally, some receptors exhibit adaptation characteristics,
39 whereby their response diminishes in the presence of continuous
40 stimuli. These receptors provide essential protection against
41 persistent and harmful stimuli while also preventing energy
42 expenditure on non-essential stimuli.
43 The volatile memristor is suitable as an artificial nociceptor because
44 it reacts only to electric pulses above a certain threshold and
45 gradually reduces the output signal once the pulse is removed.58-60
46 Moreover, such threshold and relaxation behaviors strongly depend
47 on the strength, period, and duration of the input signal. Regulating
48 relaxation enables the mimicry of phenomena observed in certain
49 receptors, such as allodynia, in which the threshold is lowered upon
50 exposure to harmful stimuli, and hyperalgesia, in which the response
51 is amplified. In addition, this approach enables the implementation
52 of adaptation functionality, which allows the receptors to adjust to
53 repeated stimuli. The detailed mechanisms and applications are
54 discussed in Section 3.
55 2.2 Neuron
56 Neurons constitute the fundamental units of the nervous system that
57 transmit electrical signals generated by external stimuli at receptors
58 in the brain, enabling recognition and response to these stimuli.61, 62
59 Neurons are primarily composed of the cell body (soma), dendrites,
60 and axons. The soma acts as the metabolic and genetic center of the
61 neuron, housing the cell nucleus and supporting vital cellular
62 functions. Dendrites extending from the soma receive signals from
63 other neurons or sensory receptors, whereas axons transmit
64 electrical signals to other neurons and muscles. These electrical
65 signals are generated from rapid changes in the membrane potential
66 of the axon, known as the action potential.63 When the action
67 potential reaches the axon terminal, neurotransmitters are released
68 into the synapse and subsequently interact with the dendrites of the
69 postsynaptic neuron. Synaptic transmission facilitates the formation
70 of complex neural networks that enable information collection,
71 integration, transmission, and coordination. Neurons are classified
72 based on their functions and characteristics. For instance, sensory
73 neurons detect external stimuli, such as light, sound, and
74 temperature, and transmit this information to the central nervous
75 system. Motor neurons carry commands from the central nervous
76 system to the muscles or glands. Interneurons function as
77 intermediaries, processing and relaying information between
78 sensory and motor neurons.
79 Volatile memristors are well-suited as artificial neurons due to their
80 ability to exhibit a steep current response exceeding a threshold
81 stimulus, followed by a decrease through volatile switching—closely
82 mimicking action potentials. Additionally, they effectively integrate
83 inputs from multiple channels and generate repetitive spike signals
84 with frequencies proportional to the combined input levels. During
85 signal generation, volatile memristors dynamically adjust their
86 responses based on input strength and frequency, efficiently
87 encoding continuous analog signals into spike trains—similar to
88 biological neurons. This adaptability enables differentiation between
89 weak and strong stimuli, replicating sensory adaptation mechanisms
90 in the human nervous system. Recent studies have demonstrated the
91 implementation of Hodgkin–Huxley (HH) and leaky integrate-and-
92 fire (LIF) model neurons using volatile memristors, further
93 highlighting their compatibility with biological neuron models. These
94 models leverage the ability of memristors to replicate essential
95 neuronal behaviors such as voltage-dependent conductance and
96 firing dynamics. Specifically, artificial neurons using volatile
97 memristors encode temporal information by adjusting their spiking
98 frequency based on the input intensity, closely resembling the time-
99 dependent stimulus information of biological sensory neurons.
100 Moreover, memristor-based implementations offer advantages such
101 as low power consumption and scalability while achieving
102 comparable performance to biological neurons.
Page 5 of 30 Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Review Journal Name
4 | J. Name., 2012, 00 , 1- 3 This journal is © The Royal Society of Chemistry 20xx
Please do not adjust margins
Please do not adjust margins
12.3 Synapse
2Synapses serve as junctions between the axon of one neuron and the
3dendrite of another, playing an essential role in neural
4transmission.52, 64 When an electrical signal reaches the axon of a
5presynaptic neuron, the synapse adjusts the connection strength
6(synaptic weight) based on the input signal, either strengthening or
7weakening the synaptic weight. The dynamic regulation of synaptic
8weight is fundamental to learning and memory and serves as a
9critical component in understanding the functional mechanisms of
10 the human brain. Adjustments in synaptic weight, such as spike-
11 timing-dependent plasticity (STDP), short-term plasticity (STP), and
12 long-term plasticity (LTP), are fundamental to the ability of the brain
13 to adapt, learn, and form memories.65-67 STDP is used to effectively
14 control synaptic weight, demonstrating a type of synaptic plasticity
15 that depends on the exact timing between the two neurons. This
16 mechanism facilitates the efficient utilization of neural networks by
17 leveraging the temporal interactions between neurons. STP refers to
18 temporary changes in synaptic strength. The STP lasts from a few
19 seconds to several minutes and can fluctuate based on the activity
20 patterns of the neurons. It is primarily governed by intracellular
21 mechanisms associated with neurotransmitter release and plays a
22 crucial role in adapting to rapidly changing environments and
23 processing transient information. Unlike STP, LTP is required for long-
24 term memory formation. LTP refers to the sustained enhancement
25 of synaptic strength over extended periods, ranging from hours to
26 years. It is known to play a critical role in learning and memory
27 processes and arises from the repeated activation of specific neural
28 paths.
29 Non-volatile memristors are highly suitable for mimicking synaptic
30 characteristics.59, 68, 69 Non-volatile memristors exhibit resistance
31 changes in response to electrical stimuli, effectively replicating the
32 synaptic weight. Furthermore, the switching behavior of non-volatile
33 memristors, which allows them to retain information even in the
34 absence of a bias, enables the emulation of long-term memory
35 functionality. The modulation of resistance and synaptic weight
36 assumes a critical function for assessing the intensity of previous
37 input signals within the frameworks of machine learning and neural
38 network algorithms. The linearity of resistance modulation is crucial
39 and can be effectively utilized to deduce the strength of the signals.
40 Linearity is essential for improving the precision of the numerous
41 algorithms used in machine learning and neural networks.
42 Furthermore, the potential of utilizing non-volatile memristors to
43 emulate the characteristics of synaptic devices has been
44 demonstrated, enabling the replication of various forms of synaptic
45 plasticity such as LTP, STP, and STDP. In detail, non-volatile
46 memristors can exhibit STDP behavior, where synaptic strength is
47 modified based on the timing of pre- and post-synaptic input spikes.
48 In addition, LTP and STP can be achieved by adjusting the device
49 conductance in response to varying input frequencies, allowing non-
50 volatile memristors to adapt to both transient and sustained input
51 patterns. This is achieved through the precise control of the
52 formation of conductive pathways, which are closely associated with
53 resistance changes in non-volatile memristors. This approach can
54 effectively reproduce the dynamic properties of synaptic plasticity.
55 These findings demonstrate the ability to implement various forms
56 of synaptic plasticity and memory functions, highlighting their
57 potential suitability for efficient brain-inspired computing
58 architectures.
59 3. Memristor-based tactile sensory system
60 Human skin enables us to recognize objects and interpret the
61 environment through the sense of touch. Tactile perception is
62 complex and involves sensing, refining, learning, and forming
63 interactions with the external environment.70-73 Receptors on
64 sensory neurons embedded in the skin, such as nociceptors,
65 chemoreceptors, and mechanoreceptors, detect various somatic
66 sensations and convey tactile information to the brain via electrical
67 signals. This process enables exquisite sensations of object
68 recognition, texture discrimination, and sensory feedback. Tactile
69 receptors can detect even small amounts of pressure or force, and
70 when combined with external stimuli, they provide a detailed and
71 nuanced picture of the object or surface being touched. This
72 information can help humans navigate their environment,
73 manipulate objects, and perform tasks that require a sense of touch.
74 They can also improve the functionality and comfort of prosthetic
75 limbs by providing users with a more natural and intuitive sense of
76 touch. This chapter explains memristor-based electronic tactile
77 sensory systems related to somatic sensations.
78 3.1 Memristor-based nociceptor and adaptive receptor
79 Nociceptors play a vital role in mimicking human acceptance and
80 processing of external stimuli. When a stimulus such as mechanical
81 stress, chemical stress, or temperature is applied, the nociceptor
82 determines the degree of hazard and generates the corresponding
83 biochemical signals. Therefore, to assess the danger posed by
84 external stimuli and to respond to and safeguard oneself, all diverse
85 features must be incorporated into the nociceptor.74, 75
86 Memristor-based nociceptors are similar to bionociceptors in that
87 they respond differently to different stimuli. As shown in Fig. 3a,
88 Yoon et al. established an artificial nociceptor based on an Ag-based
89 threshold-switching memristor with the function of a nociceptor that
90 implements four key functions (threshold, relaxation, no adaptation,
91 and sensitization).76 Allodynia and hyperalgesia, resulting from
92 harmful or abnormal stimuli, can be effectively induced in
93 memristors by applying high voltages that exceed the threshold level.
94 When the input voltage is increased to a level perceived as harmful,
95 the conductive paths in the memristor grow excessively, making
96 spontaneous and complete rupture challenging after the voltage is
97 removed. Consequently, residual Ag clusters or conductive paths
98 remain within the oxide film, facilitating a rapid response to stimuli
99 below the threshold (sensitization). To further demonstrate the
100 potential of the nociceptor, an artificial Ag-based nociceptor
101 memristor was integrated into the thermoelectric module. The
102 thermal nociceptor only generated an electric spike at a critical
103 temperature (50 °C, hazardous temperature). As the temperature
104 increased, the signal amplitude increased, and the onset time
105 decreased.
Page 6 of 30Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Journal Name Review
This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00 , 1 -3 | 5
Please do not adjust margins
Please do not adjust margins
1Kim et al.77 reported an artificial nociceptor based on a Pt/HfO2/TiN
2memristor utilizing trap/detrap mechanisms instead of a cation-
3based threshold-switching memristor. The nociceptive function was
4imitated by adjusting the trap depth of the HfO2 layer. When a
5sufficiently high positive voltage was applied to Pt, lowering the trap
6level below the Fermi energy level of TiN facilitated electron injection
7from TiN to fill the trap sites. Once filled, the electron transport
8increased sharply due to trap-assisted tunneling conduction
9between trap sites, turning the device on (threshold switching). After
10 the voltage was removed, the difference in work functions between
11 the Pt and TiN electrodes created a built-in potential that caused the
12 trapped electrons to relax over time (relaxation). The device
13 exhibited a wide operation time span ranging from milliseconds to
14 ten seconds, with a relaxation time scale well-matched to typical
15 biological systems making it highly effective for mimicking nociceptor
16 behavior. Therefore, additional circuits have been designed to
17 effectively mimic biological reflex actions, enabling immediate
18 response generation and transmission to the spinal cord when
19 exposed to danger.
20 There is an increasing need for humanoid robots to imitate advanced
21 biological functions to respond efficiently to external environments.
22 Biological skin can protect itself against harmful damage by detecting
23 the degree of danger and initiating appropriate actions using
24 nociceptors. Moreover, biological skin can self-heal and eventually
25 return to its normal state when damaged by external stimuli. The
26 design of a memristor is crucial for mimicking the complex
27 characteristics of bioskin. Xiaojie et al. reported an artificial sensory
28 system with the ability to sense and warn patients of pain and heal
29 itself. The FK-800-based organic volatile memristor acted as an
30 electronic skin (Fig. 3b).78 Self-healing was achieved because of the
31 intrinsic characteristics of the organic material, similar to human
32 skin. In addition, to sense pain and signs of injury, the artificial tactile
33 system was composed of a triboelectric generator, volatile
34 memristor, and light-emitting diode (LED). The triboelectric
35 generator and volatile switching memristor act as mechanoreceptors
36 and nociceptors, respectively. The triboelectric generator generates
37 an output voltage based on the intensity of the external stimulus,
38 and the generated voltage is applied to a volatile memristor. When a
39 voltage above the threshold value was applied to the volatile
40 memristor, the memristor and LED turned on. This case was
41 considered to have minimal damage or pain and was not considered
42 a threat. When a voltage below the threshold value was applied, the
43 memristor and LED did not turn on, causing no damage or pain.
44 Conversely, when a large input voltage was applied to the memristor
45 as a strong stimulus, the relaxation time and resistance of the volatile
46 memristor were longer and lower, respectively. Therefore, the LED
47 was stronger and required a longer time to turn off completely.
48 To effectively perceive the external environment, it is essential to
49 recognize both harmful and incoming nonharmful stimuli.
50 Nociceptors react to potentially harmful stimuli such as pressure,
51 heat, or chemicals, transmitting signals to the brain, where they are
52 interpreted as pain. They respond consistently to specific types of
53 stimuli (no adaptation). In contrast, adaptive receptors reduce their
54 sensitivity when exposed to continuous stimulation (adaptation),
55 facilitating the filtration of unimportant and repetitive information.79,
56 80 This mechanism is essential for sensory processes such as vision,
57 hearing, and touch, allowing humans to adjust to dynamic
58 surroundings.
59 However, its implementation is difficult for both the existing CMOS-
60 based and memristor-based receptors. Song et al. proposed an
61 artificial receptor that mimicked both the adaptive and maladaptive
62 characteristics using an Ag-based volatile memristor.81 The artificial
63 receptor was implemented by adjusting the thickness of the
64 conductive filament with varying amounts of metal ions. The
65 competitive relationship between Joule heating and
66 electromigration was controlled by the number of metal ions, which
67 determined the thickness of the conductive filament. Fig. 3c shows
68 that the thin conductive filament (low Ag concentration) ruptured
69 due to Joule heating during high-intensity stimuli (adaptive
70 receptor), whereas the thick filament (high Ag concentration)
71 maintained an electrical on-state (maladaptive receptor). Thus, the
72 authors demonstrated the feasibility of implementing normal
73 sensory-receptor behaviors.
74 3.2 Tactile stimulus perception
75 Artificial electronic skin, which captures surrounding tactile stimuli,
76 is deployed in advanced intelligent systems. Conventionally, artificial
77 electronic skin requires additional external equipment to store and
78 process large amounts of data. However, this structure is inefficient
79 in terms of energy consumption and processing speed because it
80 causes time delays and large energy consumption. Memristor-based
81 tactile sensory systems can effectively emulate the functions of
82 human tactile nerves in low-power operations without requiring
83 additional equipment. Memristor-based tactile sensory systems
84 enable the recording of stimuli by translating external mechanical
85 stimuli into modulated electrical spikes. To mimic a tactile sensory
86 system, an artificial system generally comprises a bio-inspired
87 synaptic or neuron memristor and various sensors for detecting the
88 external environment. The sensor connected to the memristor
89 detected the strength of the external stimulus and generated various
90 electrical signals based on the degree of stimulation applied. The
91 memristor integrates the output signals of the parallel sensor and
92 processes them into unified electrical spikes.82-84
93 Wang et al.85 demonstrated an ultrafast artificial skin system based
94 on near-sensor analog computing architecture. The artificial skin was
95 implemented by combining a memristor with a tactile sensor and
96 was fabricated on a flexible substrate. When a tactile sensor
97 recognizes an external stimulus, an input pulse is generated and
98 applied to the memristor to alter its resistance. Accordingly, the
99 system simultaneously captures and processes the tactile stimuli in
100 real time. In addition, the authors suggested that the system could
101 be mounted on a finger or prosthesis to detect the edge information
102 of external objects in real-time (Fig. 4a).
103 Sensory systems can simultaneously receive and transmit various
104 types of information from the environment via various receptors.
105 Similar to human reliance on multiple stimuli for decision-making
Page 7 of 30 Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Review Journal Name
6 | J. Name., 2012, 00 , 1- 3 This journal is © The Royal Society of Chemistry 20xx
Please do not adjust margins
Please do not adjust margins
1and responses, artificial nervous systems that utilize memristors
2require the integration of information from diverse external stimuli
3to achieve effective functionality. Artificial sensory systems aim to
4achieve multisensory functions by simultaneously integrating and
5processing various sensory input signals. The first approach involves
6integrating the input signals obtained from a circuit comprising
7multiple sensors and a memristor. Xinqiang et al.86 developed a
8multimodal sensory system that utilized pressure and temperature
9sensors in conjunction with non-volatile memristors and employed a
10 signal coupling method to integrate the outputs (Fig. 4b). The input
11 stimulus can be integrated from different sensors, and an output
12 signal can be generated once the input signal from each sensor
13 reaches a fixed threshold voltage. Six pressed and two hot stimuli
14 were applied to the system, which recognized eight stimuli and
15 generated an eight-fold output. Correspondingly, the memristor
16 reacted to several toxic stimuli and modulated conductance. This
17 study demonstrates that a multimodal artificial sensory system can
18 be constructed using different sensors (pressure and temperature)
19 and signal-coupling modules.
20 A multimodal sensory system can be realized using memristor
21 materials. This approach simplifies the circuits that constitute the
22 multimodal sensing, making it efficient and advantageous in terms of
23 energy utilization. Qingxi et al. developed a multisensory system by
24 configuring an oscillation circuit using piezoresistive sensors and a
25 VO2-based volatile memristor (Fig. 4c).87 VO2 exhibits inherent
26 thermal sensitivity, which enables its resistance state and
27 characteristics to change in response to temperature fluctuations.
28 Consequently, the VO2-based memristor enables the monitoring of
29 temperature stimuli without the need for supplementary sensors.
30 When direct thermal stimuli are applied to a memristor, the inherent
31 thermal sensitivity characteristics of VO2 alter the switching behavior,
32 thereby inducing a change in the oscillation circuit characteristics. In
33 addition, when haptic actions are applied to a piezoresistive sensor,
34 the magnitude of the stimulus alters the output of the sensor, which
35 in turn changes the voltage applied to the non-volatile memristor,
36 consequently modifying the oscillation characteristics of the volatile
37 memristor. Therefore, without multiple sensors or electrical modules,
38 an artificial mechanical sensory system can effectively synchronize
39 information regarding external stimuli through vibrations that vary
40 in response to pressure and temperature.
41 Memristor-based tactile receptors effectively detect various external
42 stimuli, including heat and pressure. These receptors mimic the
43 ability to recognize external stimulus patterns and generate
44 appropriate responses through sensor integration and
45 computational analyses. However, sensor integration remains
46 energy inefficient, and research on their ability to process multiple
47 stimuli simultaneously remains limited. Further investigation is
48 needed on software-based approaches for classifying and analyzing
49 simultaneous stimuli, such as applying algorithms similar to the
50 single-coupling module shown in Figure 4b. These additional
51 approaches can enhance the accuracy of human tactile system
52 emulation.
53 4. Memristor-based visual sensory system
54 Human vision is the primary method used to assess the size, shape,
55 color, brightness, distance, and surface roughness of an object.
56 Humans acquire more than 80% of their external information
57 through the visual sensory system. In the information acquisition
58 process, the eyes, brain, and muscles collaborate to perceive light
59 stimuli and protect oneself by responding to potentially harmful
60 stimuli.88-90 The human visual sensory system rapidly processes these
61 complicated tasks in a highly accurate and energy-efficient manner.
62 Thus, mimicking this system is desirable for the efficient detection,
63 processing, and storage of large volumes of visual information.
64 However, the biological visual system features a complex hierarchical
65 organization, including neural structures, such as the retina, bipolar
66 cells, horizontal cells, and ganglion cells. Consequently, mimicking
67 this system by using electronic circuits requires highly complex
68 circuits and substantial energy consumption for information
69 processing. Therefore, the development of more compact and
70 efficient artificial visual sensory systems that can integrate sensing,
71 processing, and storage functions is required. In Section 3, we
72 describe a method that mimics human visual characteristics, such as
73 light and motion detection, and the perception of an object using a
74 memristor. This approach employs a memristor to mimic the visual
75 adaptation functions, enhance efficiency, and reduce the complexity
76 of an artificial visual system.
77 4.1 Retina-like preprocessing
78 The retina contains photoreceptors that detect external stimuli and
79 transmit visual data to bipolar cells, which serve as intermediaries
80 between the photoreceptors and ganglion cells. The data are then
81 relayed through synapses with ganglion cells, triggering action
82 potentials that travel to the lateral geniculate nucleus (LGN). The LGN
83 transmits these signals to the visual cortex. In this process flow, a
84 memristor can process information related to light intensity, directly
85 detect the light intensity, or appropriately adapt to changes in the
86 ambient light levels of the external environment.91, 92
87 Dang et al.93 demonstrated that the one-phototransistor–one-
88 memristor (1PT1R) synaptic device shown in Fig. 5a has the potential
89 for in-sensor computing and edge computing in visual sensory
90 systems. In the 1PT1R structure, the ZnO-based phototransistor
91 provides a driving current proportional to the light illumination,
92 enabling the implementation of a high-linearity light-tunable
93 multilevel conductance state within the Mo/SiO₂/W memristor.
94 Moreover, an optical artificial neural network (OANN) composed of
95 a 16 × 3 1PT1R array performs cross-talk-free conductance updates
96 because the phototransistor functions as a selector. The proposed
97 OANN achieved a 99.3% accuracy in image recognition,
98 demonstrating that the 1PT1R device is a promising hardware
99 solution for artificial visual systems.
100 Shan et al.94 demonstrated fully light-modulated synaptic plasticity
101 using a plasmonic optoelectronic memristor comprising Ag
102 nanoparticles embedded in a TiO2 nanoporous film. Fig. 5b illustrates
103 the photooxidation and reduction processes of the Ag nanoparticles
104 embedded in the device under UV/Vis irradiation. Under visible light
105 irradiation, electrons from Ag transferred to the conduction band of
Page 8 of 30Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Journal Name Review
This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00 , 1 -3 | 7
Please do not adjust margins
Please do not adjust margins
1the TiO₂ film, generating Ag⁺ ions. This increased the effective
2diameter of the Ag conducting filament, thereby enhancing device
3conductivity. In contrast, UV irradiation excited electrons in the
4valence band of the TiO₂ film to its conduction band, which reduced
5the number of Ag+ ions and suppressed the increase in device
6conductivity. Consequently, when electrical pulses were applied
7after UV and visible-light irradiation, the current response was
8greatly improved only under visible-light irradiation. This enables the
9emulation of light-induced and gated synaptic plasticity. The STDP
10 learning was conducted using UV/Vis light. The memristor effectively
11 eliminates image noise owing to its specific UV light-induced long-
12 term depression (LTD) function. In addition, light-induced STDP
13 learning has been identified as a feature of high-level image
14 processing. By incorporating low-level image preprocessing steps,
15 such as contrast enhancement and noise reduction, the learning rate
16 and efficiency of high-level image recognition processes can be
17 significantly improved by these memristors, as demonstrated
18 through simulations.
19 Xu et al.95 reported the HH neuron-based artificial visual sensory
20 system shown in Fig. 5c using a volatile VO2 memristor. The volatile
21 VO2 memristor modulates the threshold and hold voltages based on
22 temperature, which mimics a biological neuron. The proposed
23 volatile memristor exhibits frequency relaxation in tonic spiking (a
24 type of neuron spiking model) under varying pulse inputs, and a
25 transition between spiking models when the input pulse changes
26 abruptly. This is analogous to the light-adaptive functions of
27 photoreceptors (cone and rod cells) in the retina. Primary
28 photoreceptors responsible for light processing change during the
29 transition between bright and dark environments. This shift, referred
30 to as photopic and scotopic adaptation, has been successfully
31 realized in a circuit comprising an HH neuron, a thermoelectric
32 ceramic, and a light-dependent resistor. These components convert
33 light into thermal stimuli that are subsequently used to generate
34 input pulses that induce frequency changes during spiking. This light-
35 adaptable function is useful for artificial applications. The authors
36 demonstrated the potential of integrating spiking neural network
37 (SNN) algorithms into machine vision applications to simplify circuits
38 and complex processing.
39 4.2 Self-protection via detecting the intensity of light
40 In addition to light detection, the visual system should also be
41 capable of analyzing the diverse spatiotemporal patterns of
42 photoreceptors activated in the retina. This involves protective
43 behaviors such as closing the eyes to shield against damage from
44 intense light and impending collisions, and nociceptive functions to
45 detect harmful light stimuli.
46 A highly efficient artificial visual sensory system comprising an
47 optoelectronic threshold-switching memristor and an actuator was
48 proposed by Pei et al.96 The Sb2Se3/CdS-core/shell nanorod array-
49 based (SC) optoelectronic memristor enhanced light-harvesting
50 activities, received optical signals, and converted them to a voltage
51 before transmitting them to the threshold-switching memristor-
52 based neuron circuit. The SC memristor exhibited resistive switching
53 characteristics in a light-irradiated environment, as shown in Fig. 6a,
54 driven by conductive dangling bonds and vacancy defects on the
55 surface of the Sb2Se3 nanorods. This results in an increased ON/OFF
56 resistance ratio, which in turn increases the firing frequency of
57 neuronal circuits proportional to the light intensity. When the light
58 exceeded the safety range, the firing frequency and amplitude of the
59 SC memristor and neuron circuit increased significantly, potentially
60 triggering an electric actuator. This emulates eye muscle contraction
61 and reproduces the self-protective behavior of closing eyes in
62 response to intense light damage.
63 Wang et al.97 developed an artificial visual sensory system motivated
64 by locusts, which, compared to humans, have a superior perception
65 of moving objects. The vision system of locusts includes a lobular
66 giant movement detector (LGMD) that generates danger signals
67 before the occurrence of collisions. This functionality is
68 demonstrated in Fig. 6b using an Ag conductive filament-based
69 threshold-switching memristor. The formation and rupture of Ag
70 conductive filaments in the volatile memristor were used to
71 implement the excitatory and inhibitory effects on LGMD neurons.
72 The conductivity of the volatile memristor increased and then
73 decreased as the intensity of light increased. When the light power
74 applied to the device was gradually increased to correspond to the
75 approaching objects, the current response initially increased,
76 reached a peak, and then decreased as the collision point
77 approached. In detail, at low light intensities, moderate Joule heating
78 accelerates the drift of Ag⁺ ions and the formation of conductive
79 filaments, while at high light intensities, significant Joule heating
80 induces the rupture of Ag conductive filaments. Consequently, the
81 LGMD neuron implemented in this configuration provides
82 information prior to the collision point, enabling self-protective
83 behavior.
84 Li et al.98 demonstrated a visual nociceptor based on a two-terminal
85 optical synaptic device with a monolayer MoS2 depicted in Fig. 6c.
86 The optical synaptic device successfully emulated adjustable synaptic
87 behaviors, including STP, LTP, and paired-pulse facilitation (PPF), by
88 leveraging the persistent photoconductivity resulting from charge
89 trapping. Notably, when the device was stimulated with light
90 intensities ranging from 2.5 to 7.5 nW/μm², the photocurrent
91 reached a higher level of saturation, which aligned with the no-
92 adaptation characteristic of nociceptors. Furthermore, when paired
93 320 nm light pulses were applied to the optical synaptic device at
94 intervals of 1, 2, and 3 s, a stronger photocurrent was observed at
95 shorter intervals, demonstrating the dependence of the device on
96 the relaxation time. Additionally, ultraviolet pulses with a
97 wavelength of 320 nm and power densities of 25 and 75 nW/μm²
98 were used to induce low-injured and strong-injured states,
99 respectively. In these injured states, the device exhibited a
100 heightened sensitivity to light pulses. In the low-injured state, even a
101 low-intensity ultraviolet pulse (1.5 nW/μm², 1 s) exceeded the
102 activation threshold, while in the strong-injured state, an intensity of
103 1.2 nW/μm², which is below the threshold, produced a significant
104 photocurrent. This behavior mirrors the nociceptor characteristics of
105 "allodynia" and "hyperalgesia," where sub-threshold stimuli can
106 elicit a response in an injured state.
Page 9 of 30 Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Review Journal Name
8 | J. Name., 2012, 00 , 1- 3 This journal is © The Royal Society of Chemistry 20xx
Please do not adjust margins
Please do not adjust margins
1To implement artificial visual sensory systems, memristors have been
2integrated with separate photodetection devices or fabricated using
3photoresponsive materials. While integration with separate devices
4ensures reliable processing of external stimuli, photoresponsive
5memristors offer superior integration density. However,
6incorporating photodetection capabilities into memristors often
7requires additional fabrication steps, such as coating nanorod arrays
8with photoactive materials or using ultrathin channel materials like
9nanosheets, which increases complexity. Therefore, further research
10 is required to develop simplified fabrication techniques for
11 photoresponsive memristors.
12 5. Memristor-based auditorial sensory system
13 The biological auditory system detects and collects information from
14 pressure waves of different amplitudes, frequencies, and
15 components in the medium generated by motion or collision.99-101
16 Sound waves that arrive at the ear are mechanically transmitted to
17 sensory hair cells in the cochlea, generating amplified electrical
18 signals owing to mechanical vibrations. Information in the form of
19 amplified electrical signals is transmitted from the auditory sensory
20 nerves to the cerebral cortex. Through this process, humans
21 recognize sounds in their surroundings. The input sound is encoded
22 as a train of electrical pulses created from the output of a frequency-
23 selective channel in the cochlea (space-to-rate encoding). Sparse
24 sampling of the frequency information was performed according to
25 the active frequency channel without capturing all information from
26 the sound source at the maximum sampling rate. Using this coding
27 strategy, the cerebral cortex efficiently extracts key information from
28 complex sound signals, enabling the biological auditory system to
29 produce higher-level perceptions including sound location, rhythm
30 perception, pitch recognition, and sound recognition. The ear
31 receives a combination of simultaneous sound sources with various
32 frequency components. This complexity is further exacerbated
33 because both the frequency and amplitude of these components can
34 be converted into a single sound. Owing to the spatiotemporally
35 encoded nature and time dependency of sound waves, signal
36 processing in the auditory system is more complicated than that in
37 the visual or tactile systems. Chapter 4 introduces the pioneering
38 demonstration of an integrated memristor-based artificial auditory
39 system divided into sound location (azimuth detection) and sound
40 recognition.
41 5.1. Sound location
42 To determine the location and direction of a sound source, the
43 human brain relies on interaural time difference (ITD), which is the
44 difference in the time of sound arrival between the two ears. The
45 sound signal is generally divided into a left and right signal to be
46 processed, and the important clue for sound location is the ITD in the
47 range of –0.6 ms to 0.6 ms. Based on ITD theory, several successful
48 demonstrations of sound localization have been conducted using
49 memristors.
50 To emulate sound localization based on the ITD, Sun et al.102
51 demonstrated precise temporal computation for the identification of
52 acoustic sound locations using the intrinsic synaptic capability of
53 short-term synapses. Based on the Joule heating and versatile
54 doping-induced metal-insulator transitions in a scalable monolayer
55 MoS2 device, synaptic computation was conducted to process a given
56 acoustic signal, as shown in Fig. 7a. The memristor device was
57 designed with a biologically comparable energy consumption (10 fJ),
58 and tunable STP was demonstrated by the flexible doping level of
59 MoS2. A circuit with this tunable synaptic device achieved ITD
60 detection, emulating precise temporal computations in the human
61 brain by suppressing the sound intensity- or frequency-dependent
62 synaptic connectivity.
63 The integration of piezoelectric micromachined ultrasound
64 transducer (pMUT) sensors into a neuromorphic RRAM-based
65 computational map has been reported to demonstrate real-world
66 sensory processing in object localization.103 As shown in Fig. 7b, an
67 event-driven auditory processing system applied to object
68 localization was developed using an in-memory computing
69 architecture. Inspired by the neuroanatomy of the barn owl, which is
70 known to be an efficient auditory localization system with hunting
71 capabilities during the night, the time-of-flight (ToF) of the sound
72 wave was encoded, and the difference between the two ToF
73 measurements (ITD) was analyzed to identify the sound location. The
74 energy efficiency of object localization was realized by exploiting
75 event-driven RRAM-based neuromorphic circuits that processed the
76 signal information produced by the embedded sensors to calculate
77 the position of the target object in real time. Unlike conventional
78 sensory systems that continuously sample and calculate the detected
79 signal to extract useful information, this energy-efficient auditory
80 system performs asynchronous computations as useful information
81 arrives.
82 Moreover, with the integrated 1 K HfOx-based analog memristor
83 array and a multithreshold update scheme, the in situ learning ability
84 of the sound location function was demonstrated.104 As shown in Fig.
85 7c, a brain-like learning algorithm and architecture for the sound
86 location function were successfully realized, demonstrating the
87 capability of processing sound signals from two artificial ears. With
88 high accuracy (45.7%) and energy efficiency (184×) compared to
89 existing methods, it demonstrated a significant advancement toward
90 realizing advanced auditory localization systems.
91 5.2. Speech recognition
92 Speech recognition, a key requirement for artificial intelligence
93 machines to communicate with humans, has been widely developed
94 in software-based neural networks. However, the long latency and
95 large storage requirements for large amounts of voice data in speech
96 recognition tasks in the existing von Neumann architecture pose
97 limitations. Therefore, energy-efficient neuromorphic computing
98 systems have a significant potential for processing audio signals. In
99 this subsection, several memristive-based artificial auditory systems
100 with highly accurate and efficient speech recognition performances
101 are presented.
Page 10 of 30Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Journal Name Review
This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00 , 1 -3 | 9
Please do not adjust margins
Please do not adjust margins
1A TiN/HfOx/TaOx/TiN memristor device that features a multilevel
2analog resistive state was developed.105 The artificial cochlea-based
3circuit was used to experimentally demonstrate the filtering
4behaviors of five channels with different central frequencies.
5Consequently, when connected to a convolutional neural network,
6as shown in Fig. 8a, it achieved the extraction of speech features,
7demonstrating the feasibility of a highly efficient artificial cochlear
8system.
9An artificial van der Waals hybrid synapse was developed and
10 demonstrated using acoustic pattern recognition. Its superior
11 conductance controllability was achieved using WSe2 and MoS2
12 hybrid channels, which are specialized for linear and symmetric
13 conductance change characteristics.106 The hybrid synaptic device
14 was used to perform acoustic pattern recognition (from recording,
15 transforming, and integrating) with high accuracy (93.8%), as shown
16 in Fig. 8b, indicating its potential for brain-inspired computing.
17 Speech recognition using a memristor array (W/MgO/SiO2/Mo) with
18 multilevel conductance has also been demonstrated (Fig. 8c).107
19 Speech recognition in a memristive SNN was achieved by precisely
20 tuning the weights of the artificial synapses. For effective and sparse
21 spatiotemporal feature extraction, a one-dimensional elf-organizing
22 map (SOM) network was used, which essentially operated to achieve
23 high performance and simplify the SNN classifier. Compared to other
24 ANN-based systems, the advantages of a simplified structure and
25 high energy efficiency have been demonstrated in memristive SNNs
26 for speech recognition tasks.
27 Memristors have demonstrated excellent performance in converting
28 acoustic signals into electrical signals for artificial auditory sensory
29 systems. However, a significant portion of the processing, such as
30 post-processing and learning of the converted signals, still relies
31 heavily on software-based computations and simulations.
32 Additionally, there is potential for applications that can reduce
33 sensitivity or block sounds in response to sudden loud noises, but
34 further research is needed to explore and develop these possibilities.
35 6. Memristor-based olfactory sensory system
36 The integration and coordination of the olfactory receptors, cortex,
37 and muscles enables humans to recognize and memorize odor
38 stimuli and respond to specialized gases. In the biological olfactory
39 sensory system, odorants from the environment are detected by
40 olfactory receptors, which trigger electrical signals as the output.
41 Spike signals are generated by the olfactory sensory neurons and
42 transmitted through the olfactory bulb, where signal preprocessing
43 is performed. Finally, the preprocessed signals are transmitted to
44 higher regions of the brain (olfactory cortex) to identify and
45 memorize odors.108-112 Among the various perceptions, olfaction is
46 particularly complex and vague because of the complexity of the
47 chemosensory system, which must distinguish and quantify gas
48 molecules in constantly changing environments. Therefore, these
49 olfactory processes can provide information on complex smells,
50 which in turn can provide key guidance for awareness, decision-
51 making, and action in the surrounding environment.
52 Despite the importance of the olfactory system, relatively few
53 studies have been conducted because of its complexity. It remains a
54 challenge to completely emulate the functions of the human
55 olfactory system in recognizing, memorizing, and inducing muscle
56 movements in response to dangerous gases. Section 6 introduces
57 various artificial olfactory systems based on the functions of the
58 human olfactory system, including odor recognition, memorization,
59 and protection in dangerous and gaseous environments.
60 6.1. Odor recognition and memorization
61 The olfactory system, comprising thousands of different types of
62 receptors and classifiers, enables humans to recognize and
63 memorize odors. Stimulated by odorant molecules, specific spikes
64 are generated by the olfactory receptors and analyzed using neural
65 networks. Following learning and training, humans recognize
66 different odors through memorization using olfactory systems.
67 Although various strategies have been proposed to construct
68 artificial olfactory systems, most studies have focused on developing
69 systems that use gas sensors and complex neural networks. Recently,
70 a bioinspired memristor-based olfactory system with perceptual
71 learning and memorization abilities was developed to classify several
72 different gases.
73 Qifeng Lu et al. developed a hybrid flexible gas-detection system
74 utilizing NiO nanowall-based gas sensors, oscillators, and graphene-
75 based memristor-based synapses. In this system, the signals
76 generated by the gas sensor are converted into pulses by an
77 oscillator, and the frequency of these pulses varies based on the
78 resistance of the gas sensor. The stimulation of H2S gas at various
79 concentrations was converted into pulse signals.113 The altered
80 pulses became presynaptic signals transmitted to the synaptic
81 devices, resulting in changes in the resistance (synaptic weight) of
82 the graphene-oxide-based synapse memristor. Resistance
83 modulation influences information processing and storage using
84 synaptic memristors. The system implements learning capabilities
85 based on the k-nearest neighbor (KNN) algorithm, which efficiently
86 categorizes unknown gas stimuli into the most probable categories
87 by comparing them with pre-learned boundaries. The gas-detection
88 system demonstrated enhanced recognition capabilities through
89 iterative learning. Initially, the error rate exceeded 45%; however, as
90 the number of learning iterations increased, the error rate
91 progressively decreased to approximately 20%. This methodology
92 enhances the practical application of gas-detection systems and
93 ensures reliable data analysis.
94 In addition to the mere recognition of a single gas, olfactory systems
95 have been reported to enable the detection of various gases.114 The
96 reported system utilizes an array of gas sensors along with neurons
97 and synapses to form an olfactory sensory system capable of
98 effectively analyzing complex gaseous environments. An array of gas
99 sensors capable of detecting four different gases (formaldehyde,
100 ethanol, acetone, and toluene) at various concentrations was used
101 to effectively monitor diverse gaseous environments. In a gaseous
102 environment, the resistance changes in each sensor adjusted the
103 intensity of the voltage applied to the series-connected neuronal
Page 11 of 30 Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Review Journal Name
10 | J. Name., 2012, 00, 1- 3 This journal is © The Royal Society of Chemistry 20xx
Please do not adjust margins
Please do not adjust margins
1memristor (Pt/Ag/TaOx/Pt) (Fig. 9a). These modifications to the input
2voltage translate the chemical information of the gases into electrical
3spikes in the neuron memristors, thereby providing information on
4the gas-detection capabilities of the entire system. The spikes
5generated in each neuron are transmitted to a synaptic array
6(Pt/Ta/TaOx/Pt), where they undergo learning and training through
7spike rate-dependent plasticity (SRDP). This process enables the
8storage of gas characteristics in memristor devices. Based on matrix-
9vector multiplication, the system can effectively classify four
10 different types of gases. This system enables the precise
11 identification and quantification of gases with distinct chemical
12 properties, which is highly beneficial for environmental monitoring.
13 Furthermore, these memristor-based sensory systems overcome
14 efficiency problems encountered in existing artificial sensory
15 systems, such as frequent sampling, data storage, and transfer. Han
16 et al. reported that sensors with differing sensitivities to the same
17 gas were serially connected to memristor-based neurons, proposing
18 an olfactory system capable of clearly recognizing and differentiating
19 mixed gases.115 In this system, gas exposure alters the resistance of
20 the gas sensors, modifying neuronal frequency, which can be used
21 for gas detection. Sensors based on SnO2 and WO3 exhibit different
22 resistance changes in response to the same gas, leading to distinct
23 neuronal firing frequencies. This configuration enables the artificial
24 olfactory system to distinguish unknown gases more accurately.
25 Furthermore, integration with SNNs has enhanced the ability of the
26 system to identify various types of reducing gases (NH3, CO, acetone,
27 NO2). The introduction of additional hidden layers in the SNNs
28 further improves the recognition of more complex gas mixtures,
29 highlighting its potential for environmental monitoring and safety
30 applications.
31 Currently, gas recognition and memory require additional gas
32 sensors and circuits, which adversely affect the power consumption
33 and miniaturization of the device. Chun et al. reported a system
34 capable of recognizing and remembering gases without requiring
35 additional devices or circuits by employing materials in synaptic
36 memristors that exhibited both gas-detection capabilities and
37 resistive change properties, as depicted in Fig. 9b. A synaptic
38 memristor based on Pt/TiO2 NR/TiN can directly detect gases and
39 remember them through changes in the resistance state.116 The TiO2
40 material, the oxide layer of synaptic devices, is not only used for
41 resistive switching in synaptic memristors but is also employed for
42 gas detection in conventional gas sensors. When a synaptic
43 memristor is exposed to H2 gas, the gas reacts with TiO2 to generate
44 oxygen vacancies, promoting the growth of conductive paths and
45 decreasing resistance. Conversely, exposure to NO gas removes
46 oxygen vacancies, causing disruptions in conductive paths and
47 increasing the resistance. The synaptic device detects changes in
48 resistance due to gas exposure and stores information regarding the
49 exposure. This process enables accurate recording of information
50 related to gas detection and provides reliable environmental
51 monitoring. This technology plays a crucial role in measuring and
52 managing gas concentrations in various environments. In addition,
53 the gas detection capability of a single memristor can be effectively
54 applied to mixed-gas recognition. Beyond conventional gas-sensor
55 arrays, a new approach has been reported to leverage the unique gas
56 selectivity of various materials to construct memristor arrays. This
57 study utilized SnO2, HfO2, and Ta2O5-based memristors, which exhibit
58 resistance changes in response to gas interactions. These memristors
59 demonstrated varying sensitivities to specific gases and
60 concentrations, enabling the simultaneous detection of mixed gases.
61 A parallel array significantly improved the accuracy of mixed-gas
62 concentration predictions, outperforming single-device systems by
63 over 796% compared to individual Ta2O5-based sensors. This
64 advancement underscores the potential of memristor-based sensor
65 technology to enhance environmental monitoring and improve the
66 accuracy and reliability of gas detection in complex gas
67 environments.117
68 6.2. Protection in dangerous gas environment
69 The olfactory system plays a crucial role in human awareness,
70 perception, and action in response to diverse external gaseous
71 stimuli. The coordination of olfactory receptors and muscles enables
72 humans to respond to specific gases, which is crucial for protection
73 in dangerous environments, such as in the case of leakage of toxic
74 gases or rooms on fire. However, studies on the functions of the
75 human olfactory system based on memristor devices involving
76 perception, memorization, and self-protection movements are
77 lacking. To emulate a complete olfactory system, an artificial
78 olfactory system should be developed to memorize gas information
79 and control muscles to ensure self-protection in dangerous
80 environments.
81 Recently, bioinspired olfactory systems that enable the perception
82 and memory of specific gases with the ability to act in the presence
83 of certain gases have been reported. Gas-sensing visualization using
84 a smart robot was developed for real-time gas monitoring by
85 integrating gas sensors and memory devices (Fig. 10a).118 The robot
86 was equipped with an artificial olfactory memory system developed
87 to recognize and memorize volatile organic compound (VOCs) gases
88 at different concentrations. The integration of the sensor and
89 memory unit facilitated the switching of the synaptic memristor in
90 response to the VOCs gas and recorded the target gas information
91 after the gas stimuli disappeared. Additionally, the system was
92 reconfigured with an LED to enhance the gas detection visualization.
93 When concentrations of VOCs were detected below the threshold,
94 the LED remained off. However, if the VOC concentration exceeds
95 the threshold, the LED immediately brightens and remains on. These
96 capabilities of the olfactory system present great potential for future
97 humanoid robots, environmental pollution control, and early
98 warning of chemical and biohazard safety to alert and respond to
99 emergencies in dangerous environments.
100 In addition to warning about hazardous gases, the flexible artificial
101 olfactory system shown in Fig. 10b can recognize, memorize, and
102 perform self-protection actions for NH3 and was developed by
103 integrating Sr-ZnO-based gas sensors, HfOx-based memristors, and
104 electrochemical actuators.119 The gas sensor and synaptic memristor
105 are connected in series, such that changes in NH3 concentration alter
106 the resistance of the gas sensor, which modifies the voltage intensity
107 applied to the synaptic memristor according to the voltage division
108 rule. Thus, the external chemical signals are conveyed as changes in
Page 12 of 30Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Journal Name Review
This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00, 1-3 | 11
Please do not adjust margins
Please do not adjust margins
1the electrical signals to the memristor through the resistance
2variation of the gas sensor. This process plays a crucial role in
3translating chemical stimuli into electrical signals. When exposed to
4specific concentrations of NH3, the resistance of the gas sensor
5decreased sharply; consequently, a voltage (set voltage) sufficient to
6switch the synaptic memristor was applied. When the NH3
7concentration was low, the memristor remained inactive, causing
8the actuator to remain unresponsive and the gas to flow normally.
9Conversely, as the NH3 concentration increased, the olfactory
10 memory device was activated, causing the actuator to bend inward
11 and close into a conical shape, thereby preventing gas from entering
12 the nasal cavity. Thus, the activation of the memristor triggers the
13 movement of the electrochemical actuator to block the gas flow
14 channel, mimicking the self-protective action of the induced muscle
15 movement of the hand when it smells NH3.
16 This section highlights the effective utilization of memristor-based
17 olfactory systems in humanoid robotics and environmental
18 monitoring. However, these systems face inherent limitations in
19 selectivity and sensitivity to various gases. Moreover, there is a need
20 to develop systems that can detect external gases in real time,
21 process the data, and execute appropriate responses. This approach
22 facilitates rapid and accurate reactions to gas leaks and chemical
23 hazards, significantly improving the efficiency of environmental
24 monitoring systems.
25 7. Conclusions and Perspectives
26 Memristive artificial sensory systems, inspired by the energy-
27 efficient architecture of biological systems, have been developed to
28 overcome the technological limitations of conventional CMOS-based
29 systems. Memristors can emulate the receptors, neurons, and
30 synapses—the fundamental components of biological sensory
31 systems. Building on this foundation, memristors enable higher-
32 order functions such as learning, inference, and hazard detection by
33 mimicking specific biological sensory systems. Table 1 summarizes
34 how various memristors emulate biological components and
35 implement sensory characteristics, demonstrating that memristive
36 artificial sensory systems can effectively replicate the four major
37 human senses.
38 In this review, we suggested the emulation of receptor, neuron, and
39 synapse properties using memristors based on an understanding of
40 their inherent characteristics. Volatile memristors exhibit switching
41 behavior, transitioning to an ON state when stimuli exceed a specific
42 threshold and returning to the Off state when stimuli are removed.
43 This behavior is suitable for simulating receptors and neurons. as it
44 closely resembles the "threshold" and "relaxation" responses of
45 biological receptors. In addition, by adjusting stimulus intensity and
46 duration, volatile memristors can replicate biological phenomena
47 such as adaptation and sensitization. Moreover, their behavior
48 closely resembles “the ion channel dynamics” observed in neurons.
49 When connected to an external circuit, volatile memristors can
50 effectively model spike generation, including LIF and HH models, as
51 well as neuron spike shapes. Non-volatile memristors, by contrast,
52 alter their resistance in response to an applied bias and retain their
53 resistance even after the bias is removed. This characteristic allows
54 them to mimic the information storage function of biological
55 synapses, where resistance modulation corresponds to "synaptic
56 weight" adjustments in response to neural stimuli.
57 We then discuss the implementation of the four major senses—
58 tactile, visual, auditory, and olfactory—in the memristor-based
59 artificial sensory system, as illustrated in Fig. 11. Notably, memristors
60 enable comprehensive coverage of previously unachievable
61 functionalities that play crucial roles in sensory systems and offer
62 efficient energy consumption compared to CMOS-based devices and
63 memristors (Table 2). In artificial tactile systems, advancements in
64 memristor material and structural design have enabled the effective
65 emulation of receptor characteristics such as "sensitivity" and
66 "adaptability," which were previously challenging to emulate. For
67 example, the system demonstrates a function in which the output
68 gradually decreases in response to innocuous stimuli. This
69 contradicts the conventional belief that reliable signal conversion
70 requires a consistent output for identical inputs. This aligns with the
71 operational tendencies of biological sensory systems. In the artificial
72 visual system, memristors emulate neuron-spiking models with high
73 precision to simulate the functions of biological photoreceptors. By
74 reducing the output in response to sudden increases in input signals,
75 the system facilitates "light intensity detection" and "self-
76 protection." Notably, it efficiently extracts and delivers only essential
77 information for actions, such as collision avoidance or blinking, from
78 vast visual data inputs. Furthermore, while nociceptors have
79 predominantly been implemented for tactile stimuli, the
80 development of nociceptive functionality that is responsive to visual
81 stimuli is particularly remarkable. In the artificial auditory system,
82 the memristors are connected to additional circuits that emulate the
83 filtering function of the cochlea. This system is designed to recognize
84 only specific sound amplitudes based on memristor resistance,
85 enabling "speech recognition" in the biological auditory system. This
86 represents a significant advancement in artificial auditory systems. In
87 the artificial olfactory system, memristors fabricated from gas-
88 sensitive materials integrate sensing and switching characteristics.
89 This approach allows the detection of external stimuli without an
90 additional circuit. Furthermore, memristor resistance varies
91 depending on gas type, allowing for "recognition and memorization"
92 of specific gases. These findings break the conventional stereotype
93 that receptors are solely responsible for stimulus detection while
94 synapses manage information storage. Instead, they demonstrate
95 that bioinspired and highly efficient system architectures can
96 perform multiple functions within a single device. Besides,
97 conventional CMOS-based artificial neural systems struggle to
98 implement advanced sensory functions. Even if achievable, such
99 implementations typically require significant energy consumption
100 and extended processing times. In contrast, memristor-based
101 artificial sensory systems can efficiently emulate these advanced
102 functions.
Page 13 of 30 Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Review Journal Name
12 | J. Name., 2012, 00, 1- 3 This journal is © The Royal Society of Chemistry 20xx
Please do not adjust margins
Please do not adjust margins
1While memristor-based artificial sensory systems demonstrate
2extensive potential, key challenges remain to be addressed.
3Although progress has been made in using memristors
4independently to detect stimuli and mimic sensory system functions,
5system-level integration remains challenging. Most implementations
6still rely on additional sensors and circuits primarily used for signal
7conversion, such as translating the firing frequency of artificial
8neurons into a form that other components can process. However,
9improving the energy efficiency of this conversion process has not
10 been well explored. Although memristors themselves consume nJ to
11 pJ-level low energy, integrating them with CMOS-based systems
12 often introduces mismatches in electrical parameters, requiring
13 additional circuitry for voltage conversion, signal processing, and
14 computation. This increases system complexity and overall energy
15 consumption, limiting memristors' ability to mimic biological sensory
16 systems fully. Moreover, if memristors cannot be fabricated using
17 CMOS-compatible materials and processes, chip-level integration
18 becomes extremely challenging. Without chip-level integration,
19 memristors and CMOS-based devices or circuits must be
20 implemented separately, leading to undesirable consequences such
21 as signal transmission noise, increased energy consumption, and
22 larger system area. For instance, Section 4.2 discussed a memristor-
23 based model mimicking the LGMD neuron, which was integrated into
24 a car robot to generate avoidance behavior based on optic input
25 signals. However, implementing this system required power
26 management chips for voltage conversion and counter circuits for
27 spike frequency calculation, leading to a complex structure with
28 additional energy consumption. Unfortunately, current research
29 primarily focuses on enhancing the performance of individual
30 memristor devices, with limited studies addressing CMOS
31 compatibility and efficient architectures for seamless integration
32 with CMOS-based systems. Therefore, developing a more advanced
33 memristor-based architecture is essential to enable practical and
34 energy-efficient system integration. Furthermore, addressing the
35 following challenges is imperative for the advancement of artificial
36 sensory systems. First, research on advanced data processing to
37 perform complex tasks is required. Efficient management of
38 spatiotemporal data requires multiple memristors working in
39 conjunction, along with mechanisms to compare and integrate data
40 from each device. Recent studies have primarily focused on single
41 memristors, with limited algorithms developed for arrays or circuits.
42 To mimic biological intelligence, it is essential to establish
43 interconnections among memristors and integrate their functions.
44 Additionally, research on integrated system-level memristor-based
45 receptors, neurons, and synapses is significantly lacking. To construct
46 artificial sensory systems, memristors emulate and integrate
47 receptors, neurons, and synapses. However, most studies focus on
48 them in isolation rather than as part of a cohesive system. Achieving
49 more efficient conversion and data processing between system
50 components is essential for accurately replicating biological sensory
51 functions. For artificial sensory systems to function reliably, research
52 must focus on compatible signal conversion between the pre- and
53 post-components. These investigations have the potential to
54 advance the overall integration of sensory systems by enabling
55 electrical processing of neural signals for information transmission
56 and ensuring accurate execution of output signals. In conclusion, this
57 review provides a framework for implementing memristive artificial
58 sensory systems based on the characteristics of biological
59 components and switching properties of memristors.
60 Author Contributions
61 J. E. Kim and K. Soh contributed equally to this work. J. E. Kim and K.
62 Soh conceived the review and wrote the manuscript. S. I. Hwang and
63 D. Y. Yang reviewed the manuscript. J. H. Yoon supervised the review
64 and finalized the manuscript. All authors have approved the final
65 version of the manuscript.
66 Conflicts of interest
67
The authors declare no conflict of interests.
68 Data availability
69
No primary research results, software, or codes were involved,
70
and no new data were generated or analyzed as part of this
71
review.
72 Acknowledgements
73
This research was supported by the National R&D Program
74
through the National Research Foundation of Korea (NRF) and
75
the Korea Basic Science Institute (National Research Facilities
76
and Equipment Center), and funded by the Ministry of Science
77
and ICT (RS-2024-00406418, NRF-2022R1C1C1004176, and RS-
78
2024-00403917).
79 Notes and references
80
1. P. K. Paritala, S. Manchikatla and P. K. D. V. Yarlagadda,
81
Procedia Engineering, 2017, 174, 982-991.
82
2. U. Rosolia, X. Zhang and F. Borrelli, Annual Review of
83
Control, Robotics, and Autonomous Systems, 2018, 1,
84
259-286.
85
3. Y. LeCun, Y. Bengio and G. Hinton, Nature, 2015, 521,
86
436-444.
87
4. Z. Wang, S. Joshi, S. Savel'Ev, W. Song, R. Midya, Y. Li, M.
88
Rao, P. Yan, S. Asapu, Y. Zhuo, H. Jiang, P. Lin, C. Li, J. H.
89
Yoon, N. K. Upadhyay, J. Zhang, M. Hu, J. P. Strachan, M.
90
Barnell, Q. Wu, H. Wu, R. S. Williams, Q. Xia and J. J.
91
Yang, Nature Electronics, 2018, 1, 137-145.
92
5. P. M. Sheridan, F. Cai, C. Du, W. Ma, Z. Zhang and W. D.
93
Lu, Nat Nanotechnol, 2017, 12, 784-789.
94
6. S. Seo, S. H. Jo, S. Kim, J. Shim, S. Oh, J. H. Kim, K. Heo, J.
95
W. Choi, C. Choi, S. Oh, D. Kuzum, H. P. Wong and J. H.
96
Park, Nat Commun, 2018, 9, 5106.
97
7. S. Chen, Z. Lou, D. Chen and G. Shen, Adv Mater, 2018,
98
30.
99
8. S. Kumar, R. S. Williams, Z. Wang, Nature 2020, 585, 518.
100
9. J. Wu, Y. Chua, M. Zhang, H. Li and K. C. Tan, Front
101
Neurosci, 2018, 12, 836.
102
10. H. Kalita, A. Krishnaprasad, N. Choudhary, S. Das, D. Dev,
103
Y. Ding, L. Tetard, H. S. Chung, Y. Jung, T. Roy, Scientific
104
Reports 2019, 9, 1.
Page 14 of 30Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Journal Name Review
This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00, 1-3 | 13
Please do not adjust margins
Please do not adjust margins
1
11. P. Stoliar, J. Tranchant, B. Corraze, E. Janod, M. P.
2
Besland, F. Tesler, M. Rozenberg, L. Cario, Advanced
3
Functional Materials 2017, 27.
4
12. X. Zhang, J. Lu, Z. Wang, R. Wang, J. Wei, T. Shi, C. Dou,
5
Z. Wu, J. Zhu, D. Shang, G. Xing, M. Chan, Q. Liu, M. Liu,
6
Science Bulletin 2021, 66, 1624.X. Zhang, J. Lu, Z. Wang,
7
R. Wang, J. Wei, T. Shi, C. Dou, Z. Wu, J. Zhu, D. Shang,
8
G. Xing, M. Chan, Q. Liu, M. Liu, Science Bulletin 2021,
9
66, 1624.
10
13. I. Polykretis, G. Tang, P. Balachandar and K. P.
11
Michmizos, IEEE Transactions on Medical Robotics and
12
Bionics, 2022, 4, 520-529.
13
14. W. Shi, J. Cao, Q. Zhang, Y. Li and L. Xu, IEEE Internet of
14
Things Journal, 2016, 3, 637-646.
15
15. A. Vanarse, A. Osseiran and A. Rassau, Front Neurosci,
16
2016, 10, 115.
17
16. I. T. Wang, C. C. Chang, L. W. Chiu, T. Chou, T. H. Hou,
18
Nanotechnology 2016, 27.
19
17. M. Prezioso, F. Merrikh-Bayat, B. D. Hoskins, G. C. Adam,
20
K. K. Likharev and D. B. Strukov, Nature, 2015, 521, 61-
21
64.
22
18. S. Choi, J. Yang and G. Wang, Advanced Materials, 2020,
23
32, 1-26.
24
19. D. S. Jeong, K. M. Kim, S. Kim, B. J. Choi and C. S. Hwang,
25
Advanced Electronic Materials, 2016, 2, 1-27.
26
20. H. Li, S. Wang, X. Zhang, W. Wang, R. Yang, Z. Sun, W.
27
Feng, P. Lin, Z. Wang, L. Sun and Y. Yao, Advanced
28
Intelligent Systems, 2021, 3, 2100017-2100017.
29
21. J. Park, Electronics (Switzerland), 2020, 9, 1-16.
30
22. R. Yang, H. M. Huang and X. Guo, Advanced Electronic
31
Materials, 2019, 5.
32
23. F. Cai, J. M. Correll, S. H. Lee, Y. Lim, V. Bothra, Z. Zhang,
33
M. P. Flynn and W. D. Lu, Nature Electronics, 2019, 2,
34
290-299.
35
24. J. F. Yang, A. T. Liu, T. A. Berrueta, G. Zhang, A. M.
36
Brooks, V. B. Koman, S. Yang, X. Gong, T. D. Murphey
37
and M. S. Strano, Advanced Intelligent Systems, 2022, 4,
38
2100205.
39
25. S. Wang, S. Gao, C. Tang, E. Occhipinti, C. Li, S. Wang, J.
40
Wang, H. Zhao, G. Hu, A. Nathan, R. Dahiya and L. G.
41
Occhipinti, Nature Communications, 2024, 15, 4671.
42
26. Z. Cao, L. Xiang, B. Sun, K. Gao, J. Yu, G. Zhou, X. Duan,
43
W. Yan, F. Lin, Z. Li, R. Wang, Y. Lv, F. Ren, Y. Yao and Q.
44
Lu, Materials Today Bio, 2024, 26, 101096.
45
27. M. Rao, H. Tang, J. Wu, W. Song, M. Zhang, W. Yin, Y.
46
Zhuo, F. Kiani, B. Chen, X. Jiang, H. Liu, H.-Y. Chen, R.
47
Midya, F. Ye, H. Jiang, Z. Wang, M. Wu, M. Hu, H. Wang,
48
Q. Xia, N. Ge, J. Li and J. J. Yang, Nature, 2023, 615, 823-
49
829.
50
28. X. Zhang, Y. Zhuo, Q. Luo, Z. Wu, R. Midya, Z. Wang, W.
51
Song, R. Wang, N. K. Upadhyay, Y. Fang, F. Kiani, M. Rao,
52
Y. Yang, Q. Xia, Q. Liu, M. Liu and J. J. Yang, Nature
53
Communications, 2020, 11, 1-9.
54
29. F. Li, R. Wang, C. Song, M. Zhao, H. Ren, S. Wang, K.
55
Liang, D. Li, X. Ma, B. Zhu, H. Wang and Y. Hao, ACS
56
Nano, 2021, 15, 16422-16431.
57
30. Q. Wu, B. Dang, C. Lu, G. Xu, G. Yang, J. Wang, X. Chuai,
58
N. Lu, D. Geng, H. Wang and L. Li, Nano Lett, 2020, 20,
59
8015-8023.
60
31. J. K. Han, D. M. Geum, M. W. Lee, J. M. Yu, S. K. Kim, S.
61
Kim and Y. K. Choi, Nano Lett, 2020, 20, 8781-8788.
62
32. G. Indiveri, B. Linares-Barranco, R. Legenstein, G.
63
Deligeorgis and T. Prodromakis, Nanotechnology, 2013,
64
24, 384010.
65
33. G. W. Burr, R. M. Shelby, A. Sebastian, S. Kim, S. Kim, S.
66
Sidler, K. Virwani, M. Ishii, P. Narayanan, A. Fumarola, L.
67
L. Sanches, I. Boybat, M. Le Gallo, K. Moon, J. Woo, H.
68
Hwang and Y. Leblebici, Advances in Physics: X, 2017, 2,
69
89-124.
70
34. Q. Wan, M. T. Sharbati, J. R. Erickson, Y. Du and F. Xiong,
71
Advanced Materials Technologies, 2019, 4, 1-34.
72
35. B. Joo, J. W. Han and B. S. Kong, IEEE Transactions on
73
Circuits and Systems I: Regular Papers, 2022, 69, 3632-
74
3642.
75
36. X. Yan, J. Niu, Z. Fang, J. Xu, C. Chen, Y. Zhang, Y. Sun, L.
76
Tong, J. Sun, S. Yin, Y. Shao, S. Sun, J. Zhao, M. Lanza, T.
77
Ren, J. Chen and P. Zhou, Materials Today, 2024, 80,
78
365-373.
79
37. Y. Pei, L. Yan, Z. Wu, J. Lu, J. Zhao, J. Chen, Q. Liu and X.
80
Yan, ACS Nano, 2021, 15, 17319-17326.
81
38. Y. Pei, B. Yang, X. Zhang, H. He, Y. Sun, J. Zhao, P. Chen,
82
Z. Wang, N. Sun, S. Liang, G. Gu, Q. Liu, S. Li and X. Yan,
83
Nature Communications, 2025, 16, 48.
84
39. S. Kumar, X. Wang, J. P. Strachan, Y. Yang and W. D. Lu,
85
Nature Reviews Materials, 2022, 7, 575-591.
86
40. F. Qin, Y. Zhang, H. W. Song and S. Lee, Materials
87
Advances, 2023, 4, 1850-1875.
88
41. Y. Zhang, Z. Wang, J. Zhu, Y. Yang, M. Rao, W. Song, Y.
89
Zhuo, X. Zhang, M. Cui, L. Shen, R. Huang and J. Joshua
90
Yang, Journal, 2020, 7.
91
42. S. Kumar, R. S. Williams and Z. Wang, Nature, 2020, 585,
92
518-523.
93
43. J. U. Woo, H. G. Hwang, S. M. Park, T. G. Lee and S.
94
Nahm, Applied Materials Today, 2020, 19, 100582-
95
100582.
96
44. S. R. Zhang, L. Zhou, J. Y. Mao, Y. Ren, J. Q. Yang, G. H.
97
Yang, X. Zhu, S. T. Han, V. A. L. Roy and Y. Zhou,
98
Advanced Materials Technologies, 2019, 4.
99
45. I. T. Wang, C. C. Chang, L. W. Chiu, T. Chou and T. H. Hou,
100
Nanotechnology, 2016, 27.
101
46. X. Yan, L. Zhang, H. Chen, X. Li, J. Wang, Q. Liu, C. Lu, J.
102
Chen, H. Wu and P. Zhou, Advanced Functional
103
Materials, 2018, 28, 1-10.
104
47. X. Zhang, W. Wang, Q. Liu, X. Zhao, J. Wei, R. Cao, Z. Yao,
105
X. Zhu, F. Zhang, H. Lv, S. Long and M. Liu, IEEE Electron
106
Device Letters, 2018, 39, 308-311.
107
48. H. Kalita, A. Krishnaprasad, N. Choudhary, S. Das, D. Dev,
108
Y. Ding, L. Tetard, H. S. Chung, Y. Jung and T. Roy,
109
Scientific Reports, 2019, 9, 1-8.
110
49. P. Stoliar, J. Tranchant, B. Corraze, E. Janod, M. P.
111
Besland, F. Tesler, M. Rozenberg and L. Cario, Advanced
112
Functional Materials, 2017, 27.
113
50. X. Zhang, J. Lu, Z. Wang, R. Wang, J. Wei, T. Shi, C. Dou,
114
Z. Wu, J. Zhu, D. Shang, G. Xing, M. Chan, Q. Liu and M.
115
Liu, Science Bulletin, 2021, 66, 1624-1633.
116
51. C. Yoon, G. Oh and B. H. Park, Journal, 2022, 12.
117
52. C. Wan, P. Cai, M. Wang, Y. Qian, W. Huang and X. Chen,
118
Advanced Materials, 2020, 32, 1902434.
119
53. J. Y. Kwon, J. E. Kim, J. S. Kim, S. Y. Chun, K. Soh and J. H.
120
Yoon, Exploration, 2024, 4, 20220162.
121
54. A. D. Craig, Nature reviews neuroscience, 2002, 3, 655-
122
666.
Page 15 of 30 Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Review Journal Name
14 | J. Name., 2012, 00, 1- 3 This journal is © The Royal Society of Chemistry 20xx
Please do not adjust margins
Please do not adjust margins
1
55. S. Cohen and M. E. Greenberg, Annual review of cell and
2
developmental biology, 2008, 24, 183-209.
3
56. Y. H. Jung, B. Park, J. U. Kim and T. i. Kim, Advanced
4
Materials, 2019, 31, 1803637.
5
57. J. Sandkuhler, Physiological reviews, 2009, 89, 707-758.
6
58. R. Wang, J.-Q. Yang, J.-Y. Mao, Z.-P. Wang, S. Wu, M.
7
Zhou, T. Chen, Y. Zhou and S.-T. Han, Advanced
8
Intelligent Systems, 2020, 2, 2000055.
9
59. G. Zhou, Z. Wang, B. Sun, F. Zhou, L. Sun, H. Zhao, X. Hu,
10
X. Peng, J. Yan, H. Wang, W. Wang, J. Li, B. Yan, D. Kuang,
11
Y. Wang, L. Wang and S. Duan, Advanced Electronic
12
Materials, 2022, 8, 2101127.
13
60. T. Moon, K. Soh, J. S. Kim, J. E. Kim, S. Y. Chun, K. Cho, J.
14
Yang and J. H. Yoon, Materials Horizons, 2024.
15
61. A. Vanarse, A. Osseiran and A. Rassau, Sensors, 2017, 17,
16
2591.
17
62. J. K. Han, S. Y. Yun, S. W. Lee, J. M. Yu and Y. K. Choi,
18
Advanced Functional Materials, 2022, 32, 2204102.
19
63. X. Zhang, W. Wang, Q. Liu, X. Zhao, J. Wei, R. Cao, Z. Yao,
20
X. Zhu, F. Zhang and H. Lv, IEEE Electron Device Letters,
21
2017, 39, 308-311.
22
64. G. Wei and I. H. Stevenson, Neural computation, 2021,
23
33, 2682-2709.
24
65. Q. Cheng, S.-H. Song and G. J. Augustine, Frontiers in
25
synaptic neuroscience, 2018, 10, 33.
26
66. H. R. Monday, T. J. Younts and P. E. Castillo, Annual
27
review of neuroscience, 2018, 41, 299-322.
28
67. G. Koch, V. Ponzo, F. Di Lorenzo, C. Caltagirone and D.
29
Veniero, Journal of Neuroscience, 2013, 33, 9725-9733.
30
68. T. Guo, K. Pan, Y. Jiao, B. Sun, C. Du, J. P. Mills, Z. Chen,
31
X. Zhao, L. Wei, Y. N. Zhou and Y. A. Wu, Nanoscale
32
Horizons, 2022, 7, 299-310.
33
69. S. H. Jo and W. Lu, Nano Letters, 2008, 8, 392-397.
34
70. S. Chun, J. S. Kim, Y. Yoo, Y. Choi, S. J. Jung, D. Jang, G.
35
Lee, K. I. Song, K. S. Nam, I. Youn, D. Son, C. Pang, Y.
36
Jeong, H. Jung, Y. J. Kim, B. D. Choi, J. Kim, S. P. Kim, W.
37
Park and S. Park, Nature Electronics, 2021, 4, 429-438.
38
71. D. Wang, L. Wang, W. Ran, S. Zhao, R. Yin, Y. Yan, K.
39
Jiang, Z. Lou and G. Shen, Nano Energy, 2020, 76,
40
105109-105109.
41
72. K. He, Y. Liu, M. Wang, G. Chen, Y. Jiang, J. Yu, C. Wan,
42
D. Qi, M. Xiao, W. R. Leow, H. Yang, M. Antonietti and X.
43
Chen, Advanced Materials, 2020, 32.
44
73. Y. Lee, J. Park, A. Choe, S. Cho, J. Kim and H. Ko,
45
Advanced Functional Materials, 2019, 30.
46
74. D. Dev, M. S. Shawkat, A. Krishnaprasad, Y. Jung and T.
47
Roy, IEEE Electron Device Letters, 2020, 41, 1440-1443.
48
75. M. D. Schaffler, L. J. Middleton and I. Abdus-Saboor, Curr
49
Psychiatry Rep, 2019, 21, 134.
50
76. J. H. Yoon, Z. Wang, K. M. Kim, H. Wu, V. Ravichandran,
51
Q. Xia, C. S. Hwang and J. J. Yang, Nature
52
Communications, 2018, 9, 417.
53
77. Y. Kim, Y. J. Kwon, D. E. Kwon, K. J. Yoon, J. H. Yoon, S.
54
Yoo, H. J. Kim, T. H. Park, J.-W. Han, K. M. Kim and C. S.
55
Hwang, Advanced Materials, 2018, 30, 1704320.
56
78. R. A. John, N. Tiwari, M. I. B. Patdillah, M. R. Kulkarni, N.
57
Tiwari, J. Basu, S. K. Bose, Ankit, C. J. Yu, A. Nirmal, S. K.
58
Vishwanath, C. Bartolozzi, A. Basu and N. Mathews,
59
Nature Communications, 2020, 11, 1-12.
60
79. S. Maksimovic, M. Nakatani, Y. Baba, A. M. Nelson, K. L.
61
Marshall, S. A. Wellnitz, P. Firozi, S.-H. Woo, S. Ranade
62
and A. Patapoutian, Nature, 2014, 509, 617-621.
63
80. E. L. Graczyk, B. P. Delhaye, M. A. Schiefer, S. J. Bensmaia
64
and D. J. Tyler, Journal of neural engineering, 2018, 15,
65
046002.
66
81. Y. G. Song, J. M. Suh, J. Y. Park, J. E. Kim, S. Y. Chun, J. U.
67
Kwon, H. Lee, H. W. Jang, S. Kim, C. Y. Kang and J. H.
68
Yoon, Advanced Science, 2022, 9, 1-10.
69
82. C. Wan, G. Chen, Y. Fu, M. Wang, N. Matsuhisa, S. Pan,
70
L. Pan, H. Yang, Q. Wan, L. Zhu and X. Chen, Advanced
71
Materials, 2018, 30.
72
83. C. Wang, L. Dong, D. Peng and C. Pan, Advanced
73
Intelligent Systems, 2019, 1.
74
84. Y. Lee and J. H. Ahn, ACS Nano, 2020, 14, 1220-1226.
75
85. M. Wang, J. Tu, Z. Huang, T. Wang, Z. Liu, F. Zhang, W.
76
Li, K. He, L. Pan, X. Zhang, X. Feng, Q. Liu, M. Liu and X.
77
Chen, Advanced Materials, 2022, 34, 1-8.
78
86. X. Pan, J. Wang, Z. Deng, Y. Shuai, W. Luo, W. Luo, Q. Xie,
79
Y. Xiao, S. Tang, S. Jiang, C. Wu, F. Zhu, J. Zhang and W.
80
Zhang, Advanced Intelligent Systems, 2022, 4, 2200031-
81
2200031.
82
87. Q. Duan, T. Zhang, C. Liu, R. Yuan, G. Li, P. Jun Tiw, K.
83
Yang, C. Ge, Y. Yang and R. Huang, Advanced Intelligent
84
Systems, 2022, 4, 2200039.
85
88. H. Kafaligonul, The Journal of Neurobehavioral Sciences,
86
2014, 1.
87
89. D. L. Yamins, H. Hong, C. F. Cadieu, E. A. Solomon, D.
88
Seibert and J. J. DiCarlo, Proc Natl Acad Sci U S A, 2014,
89
111, 8619-8624.
90
90. Y. Mohsenzadeh, C. Mullin, B. Lahner and A. Oliva, Sci
91
Rep, 2020, 10, 4638.
92
91. M. H. Herzog and A. M. Clarke, Front Comput Neurosci,
93
2014, 8, 135.
94
92. Yang, G. R., & Wang, X. J. Artificial neural networks for
95
neuroscientists: a primer. Neuron, 2020,107(6), 1048-
96
1070.
97
93. B. Dang, K. Liu, X. Wu, Z. Yang, L. Xu, Y. Yang and R.
98
Huang, Advanced Materials, 2022, DOI:
99
10.1002/adma.202204844.
100
94. X. Shan, C. Zhao, X. Wang, Z. Wang, S. Fu, Y. Lin, T. Zeng,
101
X. Zhao, H. Xu, X. Zhang and Y. Liu, Advanced Science,
102
2022, 9.
103
95. Y. Xu, S. Gao, Z. Li, R. Yang and X. Miao, Advanced
104
Intelligent Systems, 2022, 4, 2200210-2200210.
105
96. Y. Pei, Z. Li, B. Li, Y. Zhao, H. He, L. Yan, X. Li, J. Wang, Z.
106
Zhao, Y. Sun, Z. Zhou, J. Zhao, R. Guo, J. Chen and X. Yan,
107
Advanced Functional Materials, 2022, 32.
108
97. Y. Wang, Y. Gong, S. Huang, X. Xing, Z. Lv, J. Wang, J. Q.
109
Yang, G. Zhang, Y. Zhou and S. T. Han, Nature
110
Communications, 2021, 12.
111
98. J. Li, Y. Zhou, Y. Li, C. Yan, X.-G. Zhao, W. Xin, X. Xie, W.
112
Liu, H. Xu and Y. Liu, ACS Photonics, 2024, 11, 4578-
113
4587.
114
99. L. Ng, M. W. Kelley and D. Forrest, Nat Rev Endocrinol,
115
2013, 9, 296-307.
116
100. J. H. Lee, M. Y. Lee, Y. Lim, J. Knowles and H. W. Kim, J
117
Tissue Eng, 2018, 9, 2041731418808455.
118
101. M. Cortada, S. Levano and D. Bodmer, Int J Mol Sci,
119
2021, 22.
120
102. L. Sun, Y. Zhang, G. Hwang, J. Jiang, D. Kim, Y. A. Eshete,
121
R. Zhao and H. Yang, Nano Letters, 2018, 18, 3229-3234.
122
103. B. Gao, Y. Zhou, Q. Zhang, S. Zhang, P. Yao, Y. Xi, Q. Liu,
123
M. Zhao, W. Zhang, Z. Liu, X. Li, J. Tang, H. Qian and H.
124
Wu, Nature Communications, 2022, 13.
Page 16 of 30Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Journal Name Review
This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00, 1-3 | 15
Please do not adjust margins
Please do not adjust margins
1
104. F. Moro, E. Hardy, B. Fain, T. Dalgaty, P. Clémençon, A.
2
De Prà, E. Esmanhotto, N. Castellani, F. Blard, F. Gardien,
3
T. Mesquida, F. Rummens, D. Esseni, J. Casas, G. Indiveri,
4
M. Payvand and E. Vianello, Nature Communications,
5
2022, 13.
6
105. L. Cheng, L. Gao, X. Zhang, Z. Wu, J. Zhu, Z. Yu, Y. Yang,
7
Y. Ding, C. Li, F. Zhu, G. Wu, K. Zhou, M. Wang, T. Shi and
8
Q. Liu, Frontiers in Neuroscience, 2022, 16.
9
106. F. Moro, E. Hardy, B. Fain, T. Dalgaty, P. Clémençon, A.
10
De Prà, E. Esmanhotto, N. Castellani, F. Blard, F. Gardien,
11
T. Mesquida, F. Rummens, D. Esseni, J. Casas, G. Indiveri,
12
M. Payvand and E. Vianello, Nature Communications,
13
2022, 13, 3506.
14
107. S. Seo, B. S. Kang, J. J. Lee, H. J. Ryu, S. Kim, H. Kim, S.
15
Oh, J. Shim, K. Heo, S. Oh and J. H. Park, Nature
16
Communications, 2020, 11.
17
108. A. Rinaldi, EMBO Rep, 2007, 8, 629-633.
18
109. K. E. Whitlock and M. F. Palominos, Front Neuroanat,
19
2022, 16, 831602.
20
110. C. Huart, P. Rombaux and T. Hummel, Molecules, 2013,
21
18, 11586-11600.
22
111. G. M. Shepherd, Nature, 2006, 444, 316-321.
23
112. A. L. Saive, J. P. Royet and J. Plailly, Front Behav Neurosci,
24
2014, 8, 240.
25
113. Q. Lu, F. Sun, Y. Dai, Y. Wang, L. Liu, Z. Wang, S. Wang
26
and T. Zhang, Nano Research, 2021, 15, 423-428.
27
114. T. Wang, X.-X. Wang, J. Wen, Z.-Y. Shao, H.-M. Huang
28
and X. Guo, Advanced Intelligent Systems, 2022, 4.
29
115. J.-K. Han, M. Kang, J. Jeong, I. Cho, J.-M. Yu, K.-J. Yoon, I.
30
Park and Y.-K. Choi, Advanced Science, 2022, 9,
31
2106017.
32
116. S. Y. Chun, Y. G. Song, J. E. Kim, J. U. Kwon, K. Soh, J. Y.
33
Kwon, C.-Y. Kang and J. H. Yoon, Advanced Materials,
34
2023, 35, 2302219.
35
117. D. Lee, M. J. Yun, K. H. Kim, S. kim and H.-D. Kim, ACS
36
Sensors, 2021, 6, 4217-4224.
37
118. C. Ban, X. Min, J. Xu, F. Xiu, Y. Nie, Y. Hu, H. Zhang, M.
38
Eginligil, J. Liu, W. Zhang and W. Huang, Advanced
39
Materials Technologies, 2021, 6.
40
119. Z. Gao, S. Chen, R. Li, Z. Lou, W. Han, K. Jiang, F. Qu and
41
G. Shen, Nano Energy, 2021, 86.
42
120. D. Chen, X. Zhi, Y. Xia, S. Li, B. Xi, C. Zhao and X. Wang,
43
Small, 2023, 19, 2301196.
44
J. Huang, J. Feng, Z. Chen, Z. Dai, S. Yang, Z. Chen, H. Zhang, Z.
45
Zhou, Z. Zeng, X. Li and X. Gui, Nano Energy, 2024, 126,
46
109684.
47
123. L. Xiaoqi, J. Jianbo, L. Guangyu, Z. Bao and Z. Enming,
48
Journal of Materials Science: Materials in Electronics,
49
2024, 35, 1608
50
124. Z. Lv, S. Zhu, Y. Wang, Y. Ren, M. Luo, H. Wang, G. Zhang,
51
Y. Zhai, S. Zhao, Y. Zhou, M. Jiang, Y.-B. Leng and S.-T.
52
Han, Advanced Materials, 2024, 36, 2405145.
53
125. J. Shi, Y. Lin, Z. Wang, X. Shan, Y. Tao, X. Zhao, H. Xu and
54
Y. Liu, Advanced Materials, 2024, 36, 2314156.
55
126. Y. Gong, X. Xing, X. Wang, R. Duan, S.-T. Han and B. K.
56
Tay, Advanced Functional Materials, 2024, 34, 2406547
57
127. X. Wang, C. Chen, L. Zhu, K. Shi, B. Peng, Y. Zhu, H. Mao,
58
H. Long, S. Ke, C. Fu, Y. Zhu, C. Wan and Q. Wan, Nature
59
Communications, 2023, 14, 3444.
60
128. J. Lee, B. H. Jeong, E. Kamaraj, D. Kim, H. Kim, S. Park and
61
H. J. Park, Nature Communications, 2023, 14, 5775.
62
129. X. Shan, C. Zhao, X. Wang, Z. Wang, S. Fu, Y. Lin, T. Zeng,
63
X. Zhao, H. Xu, X. Zhang and Y. Liu, Advanced Science,
64
2022, 9, 2104632
65
130. J. Yu, F. Zeng, Q. Wan, Z. Lu and F. Pan, InfoMat, 2023,
66
5, e12458.
67
131. B. Gao, Y. Zhou, Q. Zhang, S. Zhang, P. Yao, Y. Xi, Q. Liu,
68
M. Zhao, W. Zhang, Z. Liu, X. Li, J. Tang, H. Qian and H.
69
Wu, Nature Communications, 2022, 13, 2026.
70
132. F. Moro, E. Hardy, B. Fain, T. Dalgaty, P. Clémençon, A.
71
De Prà, E. Esmanhotto, N. Castellani, F. Blard, F. Gardien,
72
T. Mesquida, F. Rummens, D. Esseni, J. Casas, G. Indiveri,
73
M. Payvand and E. Vianello, Nature Communications,
74
2022, 13, 3506.
75
133. C. Sbandati, S. Stathopoulos, P. Foster, N. D. Peer, C.
76
Sestito, A. Serb, S. Vassanelli, D. Cohen and T.
77
Prodromakis, Science Advances, 2024, 10, eadp7613.
78
134. R. Chaurasiya, K.-T. Chen, L.-C. Shih, Y.-C. Huang and J.-
79
S. Chen, Advanced Theory and Simulations, 2024, 7,
80
2301074.
81
135. L. Wang, W. Li, L. Wan and D. Wen, ACS Sensors, 2023,
82
8, 4810-4817.
83
136. T. Wang, H.-M. Huang, X.-X. Wang and X. Guo, InfoMat,
84
2021, 3, 804-813
85
137. P. Qiu, Y. Qin and Q. Xia, Sensors and Actuators B:
86
Chemical, 2022, 373, 132730.
87
138. Y. Yoon, Y. Kim, W. S. Hwang and M. Shin, Advanced
88
Electronic Materials, 2023, 9, 2300098.
89
139. Y. Qin, M. Wu, N. Yu, Z. Chen, J. Yuan and J. Wang, ACS
90
Applied Electronic Materials, 2024, 6, 4939-4947.
91
140. J.-K. Han, M. Seo, W.-K. Kim, M.-S. Kim, S.-Y. Kim, M.-S.
92
Kim, G.-J. Yun, G.-B. Lee, J.-M. Yu and Y.-K. Choi, IEEE
93
Electron Device Letters, 2019, 41, 208-211.
94
141. J.-K. Han, J. Oh, G.-J. Yun, D. Yoo, M.-S. Kim, J.-M. Yu, S.-
95
Y. Choi and Y.-K. Choi, Science Advances, 2021, 7,
96
eabg8836.
97
142. L. Wang, L. Zhang, S. Hua, Q. Fu and X. Guo, Science
98
China Materials, 2025, 1-8.
99
143. S. Seo, B. Kim, D. Kim, S. Park, T. R. Kim, J. Park, H. Jeong,
100
S.-O. Park, T. Park and H. Shin, Nature Communications,
101
2022, 13, 6431.
102
144. J. Park, Y. Jang, J. Lee, S. An, J. Mok and S. Y. Lee,
103
Advanced Electronic Materials, 2023, 9, 2201306.
104
145. X. Yan, Q. Zhao, A. P. Chen, J. Zhao, Z. Zhou, J. Wang, H.
105
Wang, L. Zhang, X. Li and Z. Xiao, Small, 2019, 15,
106
1901423.
107
146. J. Bera, A. Betal, A. Sharma, U. Shankar, A. K. Rath and S.
108
Sahu, ACS Applied Nano Materials, 2022, 5, 8502-8510.
109
147. A. Betal, J. Bera, A. Sharma, A. K. Rath and S. Sahu,
110
Physical Chemistry Chemical Physics, 2023, 25, 3737-
111
3744.
112
Page 17 of 30 Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Review Journal Name
16 | J. Name., 2012, 00, 1- 3 This journal is © The Royal Society of Chemistry 20xx
Please do not adjust margins
Please do not adjust margins
Fig. 1 Schematic of the artificial sensory system and functions, featuring integrated and collaborative networks of memristive receptors, neurons, and
synapses.
Page 18 of 30Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Journal Name Review
This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00, 1-3 | 17
Please do not adjust margins
Please do not adjust margins
Fig. 2 Features and performances required to implement artificial sensory receptors, neurons, and synapses. Function characteristics of volatile and non-
volatile memristors to mimic sensory elements.
Page 19 of 30 Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Review Journal Name
18 | J. Name., 2012, 00, 1- 3 This journal is © The Royal Society of Chemistry 20xx
Please do not adjust margins
Please do not adjust margins
Fig. 3 (a) Threshold switching behavior, allodynia, and hyperalgesia. Schematic of an artificial thermal nociceptor circuit comprising a thermoelectric module
and a volatile memristor. Generated voltage by thermoelectric module and threshold switching behavior. Reproduced with permission from ref. 76.
Copyright 2018 Springer Nature (b) Bio-inspired artificial injury response system including a sense of pain, sign of injury, and healing. Lighting of light-
emitting diodes (LEDs) according to intensity of stimulation. Reproduced with permission from ref. 78. Copyright 2022 John Wiley and Sons (c) Pulse
response of memristors to multiple 100 µs pulse widths with an amplitude of 3 V. Adaptation rates of 1, 2, and 3 nm Ag memristors are classified as rapidly,
slowly, and no-adapting, respectively. Circuit schematic of an artificial sensory nervous system. Generated voltage from the thermoelectric module and
volatile memristors was monitored by oscilloscope channels at hot plate temperatures of 40 and 70 °C. Reproduced with permission from ref. 81. Copyright
2021 John Wiley and Sons
Page 20 of 30Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Journal Name Review
This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00, 1-3 | 19
Please do not adjust margins
Please do not adjust margins
Fig. 4 (a) Near-sensor analog computing using artificial tactile system. Resistance changes in synapse memristor using a continuous pulse train. Near-sensor
analog computing for real-time edge detection of the captured pressure pattern. Reproduced with permission from ref. 85. Copyright 2022 John Wiley and
Sons (b) Multimodal sensory system with multi sensors accepting pressure and temperature stimuli. Resistance modulation of the pressure and temperature
sensors as a response to pressure and hot stimuli. Reproduced with permission from ref. 86. Copyright 2022 John Wiley and Sons (c) Characterization of
artificial temperature perception VO2-based neuron memristor. Haptic-temperature fusion is based on a VO2 volatile memristor and MLP by simulation.
Reproduced with permission from ref. 87. Copyright 2022 John Wiley and Sons
Page 21 of 30 Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Review Journal Name
20 | J. Name., 2012, 00, 1- 3 This journal is © The Royal Society of Chemistry 20xx
Please do not adjust margins
Please do not adjust margins
Fig. 5 (a) Schematic illustration of the integrated 1PT1R structure device and light-tunable conductance update performance of the device. Reproduced with
permission from ref. 93. Copyright 2023 John Wiley and Sons (b) Schematic illustration of light-induced synaptic modification mechanism based on photo-
induced redox reaction and current response after UV/Vis light irradiation. Reproduced with permission from ref. 94 Copyright 2021 John Wiley and Sons (c)
Bio-inspired HH neuron for artificial retinal system with firing frequency modulated in a manner similar to photopic/scotopic adaptation of a biological
photoreceptor. Reproduced with permission from ref. 95. Copyright 2022 John Wiley and Sons
Page 22 of 30Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Journal Name Review
This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00, 1-3 | 21
Please do not adjust margins
Please do not adjust margins
Fig. 6 (a) Multifunctional artificial visual perception nervous system using optoelectronic memristor based on an Sb2Se3 nanorod array. Increasing ON/OFF
resistance ratio under light irradiation increases the firing frequency, activating an eyelid-shaped actuator. Reproduced with permission from ref. 96.
Copyright 2022 John Wiley and Sons (b) Schematics of the artificial LGMD neuron device and current response under looming light stimulus. The formation
of the Ag conductive filament is initially facilitated by the increasing light stimulus but ruptures due to Joule heating beyond a certain light intensity,
providing information before the collision point. Reproduced with permission from ref. 97. Copyright 2021 Springer Nature (c) Schematic of the monolayer
MoS2 device and current response under varying light intensity, pulse interval, and degree of injury. Reproduced with permission from ref. 98. Copyright
2024 American Chemical Society
Page 23 of 30 Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Review Journal Name
22 | J. Name., 2012, 00, 1- 3 This journal is © The Royal Society of Chemistry 20xx
Please do not adjust margins
Please do not adjust margins
Fig. 7 (a) Schematic of the human auditory perception system and monolayer MoS2-based device with Joule heating-driven conductance facilitation. ITD-
based sound localization can be achieved by suppressing interference and encoding only ITD information through artificial synaptic computation comprising
the MoS2 device. Reproduced with permission from ref. 102. Copyright 2021 American Chemical Society (b) Object localization system in barn owls and
proposed bio-inspired technology. Response varies across population, impacting both input gain and time constant. Owing to neuron-to-neuron variability,
two output neurons of direction-sensitive coincidence detector respond differently to input stimuli. Thus, sound source can be identified. Reproduced with
permission from ref. 103. Copyright 2022 Springer Nature (c) Conceptual diagram of memristor-based neuromorphic sound localization system. Multiple
binaural features applied for neural processing to detect sound sources, including binaural time difference, spectral shape, etc. Reproduced with permission
from ref. 104. Copyright 2022 Springer Nature
Page 24 of 30Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Journal Name Review
This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00, 1-3 | 23
Please do not adjust margins
Please do not adjust margins
Fig. 8 (a) Schematic of artificial cochlea speech recognition system used to demonstrate frequency-selection function of five channels in the cochlea.
Channels have central frequencies determined by the resistance of a memristor. It achieved a recognition accuracy of 92% using 64 channels. Reproduced
with permission from ref. 104. Copyright 2022 Frontiers Media S.A. (b) Design procedure of acoustic pattern (from recording, through transforming, to
integrating). The van der Waals hybrid synapse was utilized to perform acoustic pattern recognition, a common task in speech and sound processing.
Reproduced with permission from ref. 105. Copyright 2020 Springer Nature (c) Schematic of feature extraction from speech signals. Extracting features from
speech signals enables successful training of SNN in both software- and memristor-based implementations, resulting in accurate classification inference.
Reproduced with permission from ref. 107. Copyright 2021 John Wiley and Sons
Page 25 of 30 Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Review Journal Name
24 | J. Name., 2012, 00, 1- 3 This journal is © The Royal Society of Chemistry 20xx
Please do not adjust margins
Please do not adjust margins
Fig. 9 (a) Bio-inspired neuromorphic olfactory system based on the memristive neural network comprising a gas sensor, sensory neurons, synapse arrays,
and relay neurons. Sampling voltages in the LIF neuron. Larger input signals (red lines) results in shorter capacitor charging times (green lines), quicker
device switching (blue lines), and higher output frequencies (orange lines). Training loss and testing accuracy of detection gas. Reproduced with permission
from ref. 103. Copyright 2022 John Wiley and Sons (b) Schematic of biological olfactory cognitive process mimicking using chemi-memristive sensor.
Response curves upon exposure to 1% H2 and I–V curves of TiO2 NRs. Conductance modulations based on type of target gas (reducing or oxidizing).
Reproduced with permission from ref. 104. Copyright 2023 John Wiley and Sons
Page 26 of 30Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Journal Name Review
This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00, 1-3 | 25
Please do not adjust margins
Please do not adjust margins
Fig. 10 (a) Sensory information provided by volatile organic compounds sensed by olfactory sensory receptors. Demonstration of robot equipped with
artificial olfactory memory system to visualize gas sensing. Higher concentration of VOCs above threshold resulted in switching of memory device and
lighting up of LED. Reproduced with permission from ref. 118. Copyright 2021 John Wiley and Sons (b) Schematic illustration of nose comprised of four arc
actuators. Response of the bionic nose to high concentration (500 ppm) ammonia and instantaneous current changes of artificial olfactory system and
actuator array. Reproduced with permission from ref. 119. Copyright 2021 Elsevier
Page 27 of 30 Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Review Journal Name
26 | J. Name., 2012, 00, 1- 3 This journal is © The Royal Society of Chemistry 20xx
Please do not adjust margins
Please do not adjust margins
Fig. 11 Schematic of biological and artificial sensory systems with a memristor.
Page 28 of 30Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Journal Name Review
This journal is © The Royal Society of Chemistry 20xx J. Name., 2013, 00, 1-3 | 27
Please do not adjust margins
Please do not adjust margins
Table 1. A summary of memristive artificial sensory systems.
Sense
Memristor
Materials & Structure
Biological Counterpart
Specific feature
Ref
Non-volatile
Ag/CsPbBr3/PVA/FTO
Synapse
Mechanoreceptor
(Pressure)
120
Non-volatile
Al/CS:MWCNTs/ITO
Synapse
Mechanoreceptor
(Pressure)
121
Non-volatile
Ag/TiOx/Ti3C2Tx/Au
Neuron
Mechanoreceptor
(Pressure)
122
Volatile
Al/ZnO/FTO
Synapse
Nociceptor
123
Tactile
Volatile
Ag/c-YY NW/Ag
Neuron
Mechanoreceptor
(Huminity)
124
Volatile
Al/Ag NW-embedded SA/SA/ITO
Synapse/Neuron
Scotopic
/photopic adaptation
125
Volatile
Cr/Au/WS₂/Cr/Au
Synapse
Color recognition
126
Volatile
ITO/Ta2O5/Ag/IGZO/ITO
Neuron
Color recognition
127
Volatile/non-volatile
FTO/NiO/Organic Interlayer/PMMA/Ag
Synapse
Color recognition
128
Visual
Volatile/non-volatile
Pd/TiOx/ZnO/TiN
Synapse
Object tracking
129
Volatile
Pd/Nb/NbOx/Nb/Pd
Synapse
Sound Localization
130
Non-volatile
TiN/TaOy/HfOx/TiN
Synapse
Sound Localization
131
Non-volatile
TiN/HfOx/Ti/TiN
Synapse
Object localization
132
Auditory
Non-volatile
Pt/TiOx/AlOx/Pt
Synapse
Audio-
Reward association
133
Non-volatile
Ta/m-ZrO₂/Pt
Synapse
Odor recognition
134
Non-volatile
Al/pectin:Ag NPs/ITO
Synapse
Odor recognition
135
Volatile/non-volatile
W/WO₃/PEDOT:PSS/Pt, Pd/W/WO₃/Pd
Synapse/Neuron
Gas-Classification
136
Olfactory
Non-volatile
-/TiO₂ Nanowire/ Ti
Sensor
Odor recognition
137
Page 29 of 30 Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F

Review Journal Name
28 | J. Name., 2012, 00, 1- 3 This journal is © The Royal Society of Chemistry 20xx
Please do not adjust margins
Please do not adjust margins
Table 2. Comparison of switching characteristics with CMOS-based devices.
Structure
Operating voltage
Switching Speed
ION
Ref
CMOS
Sn-doped polycrystalline 𝛽-
Ga2O3 FET
10 V (VD)
0.5 s
-
138
Ag/SnSe/Au
Set : 0.474
55 ns
10 μA
139
Receptor
Memristor
Pt/Ag/SiO₂ NRs/Ag/Pt
Set : - 0.72 V
+ 0.78 V
20 μs
1 μA
81
Si-based MOSFET
VG : -1 V
VD : > 3.5 V
0.1 s
≈
150 μA
140
CMOS
Si/SiO2/Si3N4/SiO2/Si-based
MOSFET
VG : 12 V
VD : > 3 V
0.02 s
≈
150 μA
141
Pt/Ag/TaOx/Pt
Set : 0.29 V
80 μs
0.1 μA
114
Ag/MoS2 nanosheet/
Ag/MoOx/Ag
Set : 0.3 V
16 ns
100 μA
31
Neuron
Memristor
Pt/Ag/ZnO/Pt
Set : 0.17 V
≈
50 ns
10 μA
142
Si/WOx/SiO2-based FET
Write: 1.8 V (VG)
Erase : -2.5 (VG)
0.3 ms
-
143
CMOS
IGZO channel-based FET
Write: 20 V (VG)
Erase : -20 V (VG)
100 ms (Write)
10 ms (erase)
≈
10 μA
144
Pd/WS2/Pt
Set : 0.6 V
Reset : -0.2 V
14 ns
1 μA
145
Al/PVP-CdSe QD/Al
Set : 0.61 V
Reset : -0.5 V
41 ns
5.2 μA
146
Synapse
Memristor
ITO/CdS QDs-PVP/Al
Set : 1.08 V
Reset : -0.72 V
42 ns
4.44 μA
147
Page 30 of 30Materials Horizons
Materials Horizons Accepted Manuscript
Open Access Article. Published on 07 March 2025. Downloaded on 3/7/2025 12:15:13 PM.
This article is licensed under a
Creative Commons Attribution-NonCommercial 3.0 Unported Licence.
View Article Online
DOI: 10.1039/D5MH00038F