A Review on Magnetic Smart Skin as Human–Machine Interfaces

In recent years, there is a meteoric rise in the prevalence of electronic wearables, and flexible wearables have attracted tremendous attention in human–machine interfaces due to their high biocompatibility, functionality, conformability, and low‐cost. Flexible magnetic smart skin is part of this rapidly progressing field of flexible wearable electronics, which has paved the way for a new path for human perceptual development, as they may form a new perception, also known as magnetic sense, that is, the capacity to detect and interact with magnetic fields. In this review, the concept of the magnetic smart skin is defined: magnetoelastomers and flexible magnetic sensors. In magnetoelastomers, different sources of magnetic field, structures, and applications are summarized by recent and renowned research. In flexible magnetic sensors, different flexible substrates, sensor types, and applications are also summarized. This review further concludes by discussing the outlook and current challenges of magnetic smart skin in human–machine interfaces.


Introduction
[3][4] These electronic devices have a broad variety of uses, some of which include monitoring the health state of humans, [1,2,[5][6][7] tracking the movement and activities of human beings and functioning as a human-to-machine DOI: 10.1002/aelm.202300677 In the not-toodistant future, electronic gadgets will be unnoticeable and will build a connection that is seamless between the analog and digital worlds.As a result, it is necessary for electronic equipment to have greater flexibility, even reversible tensile strain.The electronic skin is a multi-functional technology that was inspired by nature's archetype.16][17][18][19] The electronic skin has become a technology that can be used to restore or enhance human perception.[22] On the other hand, the electronic skin is capable of imitating functions that are already possessed by humans.
Magnetic smart skin is part of the rapidly progressing field of flexible electronics, which has brought forward different types of magnetoelastomer, [23][24] sensors (such as flexible magnetic tunnel junctions, flexible giant magneto-resistance sensors, and flexible hall sensors) [25,26] and magnetic skins. [24,27]These magnetic smart skins pave the way for a new path in human perceptual development, as they may form a new perception, also known as magnetic sense, that is, the capacity to detect and interact with magnetic fields.This opens up new avenues for research into how humans perceive the world.The magnetic smart skin consists of two components: the magnetoelastomer and the flexible magnetic sensor.The magnetic smart skin can be attached easily to the human due to the flexibility and work as the humanmachine interfaces.][30] Moreover, the magnetic field can be utilized to monitor 3D movement, which opens up new possibilities for the control of interactive gadgets using a variety of hand gestures.This tracking method can partially replace the optical system because, when the sight is disturbed, the system is unable to easily capture the subtle movement, but the magnetic field does not have these restrictions.
The use of ambient or human-generated magnetic fields to facilitate communication between humans and computers is the topic of this article.With magnetic smart skin, humans can create magnetic fields or sense their presence.As a result, the evaluation of magnetic smart skin for HMI purposes has to be split into two sections: the first deals with magnetoelastomer, while the second discusses flexible magnetic sensors.In the context of virtual and augmented reality, a recent study, [31] briefly cites certain instances of magnetosensitive e-skins, however it only analyzes the characteristics of the flexible magnetic sensors.The magnetoelastomer, another key component, is ignored.Incorporating the subject matter is crucial to provide a more comprehensive picture of the field under question.
Here, we will review the magnetic smart skin as humanmachine interfaces in the recent years.In the interfaces, the magnetoelastomer can be used as a magnetic field source, and flexible magnetic sensors can be used as a transducer.Together, they have made great contributions to the magnetic human-machine interface.The structure of the review is as follows.In Section 2, we provide a broad overview of the different materials and structures that are currently being investigated for magnetoelastomer, and we also summarize the relevant applications of magnetoelastomer in human-machine interactions.Thereafter in Section 3, this review describes the flexible magnetic sensors, in which various substrates and sensor types as well as applications are assessed.In the final section, we summarized the ways in which magnetic smart skins will likely be developed and enhanced in the near future (Figure 1).

Magnetoelastomer
Human beings are capable of generating magnetic fields, such as magnetocardiogram, magnetoencephalography, and biomagnetic.But these spontaneous low-frequency magnetic fields of the human are too weak to be a stable input for magnetic field perception in the HMI.However, magnetic elastomers, which are created by embedding magnetic compounds in elastomers, can assist people in becoming sources of magnetic fields.The magnetoelastomer offers extreme flexibility and stretchability, whilst being lightweight and adding tunable permanent magnetic properties.It is pleasant and invisible to wear due to the material's pliability, which eliminates the feeling of its existence when linked to the body.These undetectable magnetoelastomer enable gesture control in a variety of forms and provide remote operation options for a wide range of applications. [27,32]ypically, the magnetoelastomer is composed of an elastic matrix and permanent magnetic substances.The elastic matrix materials can be PDMS or Ecoflex. [33]PDMS is a kind of silicone that is regarded as inert, non-toxic, and non-combustible, whereas Ecoflex is a very soft and transparent silicone rubber.The viability of cells growing on PDMS and Ecoflex remains very high, as evaluated using the PrestoBlue cell viability assay and the LIVE/DEAD fluorescence staining method, which indicated the great biocompatibility of the two materials. [27]n the next section, we shall discuss separately the development of the source of magnetic field, the structure of magnetoelastomers, and their applications.

Source of Magnetic Field
The magnetism of matter takes into account the intrinsic magnetic moments of elementary particles.In the case of magnetic compounds in the magnetoelastomer, for instance, a significant number of studies have made use of magnets as the driving force of the magnetic field.The magnetism for magnets originates mainly from the magnetic domains formed by the neat arrangement of electron spins.However, in addition to magnets, magnetic field sources also include permanent magnetic particles and coils, the magnetic fields are generated from the neatly arranged magnetic domains after magnetization and the induced magnetic fields after energization, respectively.

Magnets
Magnetic fields can be generated in a variety of ways, with the most common of which is by using magnets.At present, some components of permanent magnet materials, such as magnetite, are obtained from the natural environment. [34]On the other hand, as technology has progressed, there has been an increase in the usage of artificial magnets, such as ferroalloys that have been doped with components of aluminum and nickel. [35]In addition to naturally occurring magnets, the application of artificial magnets has grown increasingly common.In particular, rareearth magnets are the ones that are used the most frequently used in magnetoelastomers (such as NdFeB and SmCo magnets), the rare earth magnets have the advantages of high remanence, large coercive force, and uneasy to demagnetize compared to other magnets. [36]The above are basically hard-magnetic materials, the most distinctive feature of hard-magnetic materials is that they can become a stable and permanent source of magnetic field after being magnetized, due to their large magnetic hysteresis characterized by the high coercivity and high remanence.For example, the coercivities of alnico (iron alloys consisting of aluminum, nickel, cobalt, and copper) range from 0.4 to 2 kOe, and it can even exceed 12.5 kOe in SmCo or NdFeB. [37]The high coercivity allows them to be relatively insensitive to external conditions (e.g., temperature and magnetic fields) (Figure 2a).
Since the amplitude and direction of the magnetic field that is created by the magnet are both relatively stable, it is simple to make adjustments to any variable applications.In addition, the magnetic field's intensity is far higher than that of t some stray magnetic fields, such as the geomagnetic field (≈0.5 Oe).As a result of this, it is possible to lessen the adverse effects that are brought on by the presence of stray magnetic fields.In 2017, Sunjong Oh et al. initiated a highly sensitive and robust tactile sensing system inspired by the human synaptic system. [38]External pressure on the remote touch tip is transferred in the form of air pressure to the magnetic synapse, which is composed of a magnet.They also demonstrated applications of the remote tactile sensor, namely surface texture discrimination, heartbeat measurement, and satisfactory detection quality in water, show-ing that it has considerable potential for use in robotics, medical diagnosis, and prosthetics.
The magnets will continue to hold their position as the source of magnetic field generation caused by the advantages on the stable and great magnetic field, and there are lot of research that uses magnets to provide a magnetic field, [39][40][41][42][43][44] and they will still take its place in the source of magnetic field generation in the future.However, the magnets used to create magnetoelastomers have a high volume, which will have a negative impact on the device's ability to function as a wearable magnetic field source.This goes against the lightweight, flexibility, and stretchability of magnetic smart skin, limiting its usefulness to situations where flexibility is less of a need, and it can only be used in certain robotic setups and is not human-friendly.

Magnetic Particles
As previously mentioned, the mechanical property degradation of magnetoelastomers is inevitably caused by the permanent magnet.Researchers are looking at the possibility of using magnetic particles as a means of enhancing the mechanical properties of a material.Magnetic particles are similar to magnets in that they both fall under the category of permanent magnets.However, the key difference lies in the fact that magnets are inherently rigid, generating fixed and unchanging magnetic fields.To change the magnetic field produced by a magnet, the only option is to replace the magnet itself.In contrast, magnetic particles can be dispersed within an elastic matrix, allowing the magnetic field to change in response to the deformation of the elastic matrix.This flexibility makes magnetic particles more promising for applications in the field of human-machine interaction compared to rigid magnets (Figure 2c).
The materials of permanent magnetic particles are usually Nd-FeB or FeO, and the diameter can be as low as several microns or even several tens of nanometers.The smaller size of the magnetic particles in the magnetic elastomer can lead to a significant improvement in the material's mechanical characteristics.In addition, the researchers changed the material of the elastic matrix (Ecoflex (Smooth-on, 00-50) or hydrogel) to further improve the stretchability of magnetoelastomers. [33,46]Still, the materials of the elastic matrix are not related to the point of this review, and here we will not be broadly addressed.
[48][49][50][51][52][53][54] The following is the general process of making magnetoelastomers by magnetic particles: the first step is to mix the elastic matrix with permanent magnetic particles to prepare the composite material, and then the 3D print mold is prepared with the required shape and size.The mixture is cast into a mold, and a casting blade is used for planarization.After curing, the magnetic skin is magnetized in the out-of-plane direction.Finally, the composite material was released from the mold to form magnetoelastomers.
The quantitative analysis between the magnetic skin properties and the materials and its ratio in Ref [33] indicated that an increase of remanence is caused by the increase in the proportion of magnetic powder (Figure 2d,e).Moreover, the remanence is not related to the material of the elastic matrix, but only related  [41] Copyright 2016, The Authors, published by MDPI.b) Coils.Reprinted with permission from. [45]Copyright 2020 American Chemical Society c) Magnetic particles.Reproduced under the terms of the CC-BY Creative Commons Attribution 4.0 International license. [39]Copyright 2018, The Authors, published by MDPI.d,e) The characterization of the modulus of elasticity and remanence versus different content of the magnetic powder in PDMS and Ecoflex.Reproduced under the terms of the CC-BY Creative Commons Attribution 4.0 International license. [33]Copyright 2022, The Authors, published by Wiley-VCH GmbH.
to the proportion of magnetic powder by comparing the remanence of the composite materials when PDMS and Ecoflex are used as the elastic matrix.Besides, compared with Ecoflex, the composite material with PDMS as the elastic matrix has a larger young's modulus and worse stretchability.As a result, Ecoflex is often preferred in certain magnetic thin films, while the larger elastic modulus of PDMS is better suited for magnetic cilia.Balancing remanence and elastic modulus are essential for different applications.
At present, this magnetoelastomer has also been studied in the human-machine interaction.In 2021, Abdullah S. Almansouri et al. developed a comprehensive assistive magnetic skin system that allows quadriplegics, including the severely injured ones, to move around individually and control their surroundings with ease. [54]In 2020, Tess Hellebrekers et al. fabricated a magnetoelastomer for single-point contact localization and force estimation. [53]In 2022, Zhang et al. initiated a system for gesture recognition and magnetic positioning. [33]In 2021, Zhang et al. developed a novel magneto-piezoresistive proximity sensor by having a hierarchical magnetoelastomer coated with a 3D piezoresistive network for touchless tactile perception. [55]et, there are several downsides associated with magnetic particles.Because the magnetic field created by magnetic particles is relatively weak and limited compared to that generated by a permanent magnet, the working distance that it can cover is significantly reduced.In addition to this, the magnetic powder requires further magnetization, which adds another step to the process and results in an increase in both the production and the time cost.

Other Magnetic Substances
In addition, coils are also a traditional way to generate magnetic fields except for the magnetic particles and permanent magnets.Nevertheless, the coil is rather huge and has a number of restrictions when used as a wearable device.However, thanks to the magnetic flux of the coil, it is also very applicable as an energy supply to wearable devices by generating electricity through a magnetic field.The magnitude of the induction potential is related to the speed of change of the magnetic flux through the closed circuit, as Equation (1) shows: The equation solves for the power supply in terms of the number of turns(n) and the magnetic flux () changing with time(t).
In 2020, Edward J. Barron et al. demonstrated magnetic colloidal suspensions as compliant fluid inclusions in an elastomer matrix leading to exceptional magneto-mechanical properties for use in wearable wireless power transfer systems. [45]The magnetic permeabilities were measured to be nearly identical between the rigid composite and suspension composite samples at the same composite volume fraction, leading suspension composite samples are excellent candidates for use in wearable inductor backplanes for wireless power transfer (Figure 2b).
Maintaining a proper power supply is a crucial consideration in HMI devices.Traditional tethered and battery power suffer from motion restrictions and sophisticated recharging processes, respectively, which introduce soft-hard interfacial instability and are obstacles to the application of wearable and implantable devices.Coils demonstrated these composites for applications in wearable electronics by creating soft, magnetic backplanes for stretchable inductors, which enable wireless power transfer to be integrated into consumer clothing.The technologies based on electromagnetic energy transmission can be used to harvest ambient electromagnetic energy emitted from power lines and electronic devices, leading to an energy recycling effort for sustainability, but it has relatively low power density and recoverable power (10 nW−100 mW, [56,57] ).However, by bringing the field source and effectors in close proximity, the near-field magnetic soft machines have offered appealing applications in portable electronics. [58,59]Hence, electromagnetic energy could be a good add-on, and there should be more research done on this in the future.

Structure of Magnetoelastomers
In addition, the Magnetoelastomer structure is also interesting to investigate in HMI.Human-machine interaction applications can benefit greatly from a sensitive framework that is tailored to their specific needs.And an appropriate sensitive structure can effectively enhance the application performance.The common structures in the HMI are magnetic film, magnetic cilia, and other microstructures.

Magnetic Film
Typically, a magnetoelastomer will take the form of a magnetic film.Magnetic film is widely employed in current research for direct human-machine interfaces.Combining the features of flexibility, stretchability, and biocompatibility, along with versatility in shape and color, makes the magnetic film imperceptible to wear (Figure 3a).In light of this, the magnetic film structure is commonly found in wearable electronics.Such as the estimation of location points, three-dimensional force magnitude, and direction.In 2016, Tito Tomo et al. took the lead in using magnetic films to estimate contact positions. [41]In 2019, Tess Hellebrekers et al. used magnetic films to estimate forces magnitude and positions, [52] and then further used neural networks to improve estimation resolution and performance in 2020. [53]In 2022, Yan et al. proposed a novel tactile super-resolution method based on a sinusoidally magnetized soft magnetic film, by which we have achieved a 15-fold improvement in localization accuracy (from 6 to 0.4 mm) as well as the ability to measure the force magnitude.Film structures are more suitable for large-area tactile sensing, but it is difficult to eliminate the force direction. [50]However, in 2021, Yan et al. reported a soft tactile sensor with both selfdecoupling and super-resolution abilities, which can distinguish the normal force and the tangential force. [48]he above work is all about the force and contact point estimation.However, the magnetic films formed different shapes when contacting different objects, and the film structure can also be used for texture and object recognition.In 2022, Lin et al. developed a new approach through magnetic films to enable surgeons to preview the post-operative effects of an artificial disc implant in a patient-specific fashion prior to surgery. [4]igure 3. Three different structures of magnetoelastomer: a) Magnetic film.Reproduced with permission. [27]Copyright 2019, Wiley-VCH GmbH.b) Magnetic cilia Reproduced with permission. [79]Copyright 2015, Wiley-VCH GmbH.c) Magnetic pyramid-shaped surface.Reproduced under the terms of the CC-BY Creative Commons Attribution 4.0 International license. [46]Copyright 2019, The Authors, published by Springer Nature.
In addition, the film structure can also be used for noncontact human-machine interaction, the magnetic film can be easily attached to the human which is different from other forms.Also, the utilization of magnetic films in this field has grown significantly and has become a major direction in recent years.For example, in 2018, Almansouri et al. proposed an imperceptible magnetic skin for contactless HMI, [27] after that, many works used the magnetic film for wheelchair control, [54] eye movement recognition, [40] and catheter tracking. [60,61]The film structures focus on the large-area sensing and contactless HMI.

Magnetic Cilia
Cilia are slender protrusions of cells, which are ubiquitously present in nature, acting both as actuators and sensors. [62]Inspired by nature, scientists have developed artificial cilia mimicking the functions of biological cilia.Artificial cilia research has been ongoing for more than a decade, and it is still rapidly expanding.Many research groups have developed microscopic actuators resembling cilia, actuated to move under the influence of different stimuli such as electrostatic fields, [63] magnetic fields, [64][65][66][67][68][69] and even light [70] and pH. [71]As compared to the film structure, the magnetic cilia provide a significant gain in detection limit and sensitivity, often by a factor of hundreds.
Since 2014, Alfadhel Ahmed and Pedro Ribeiro have successively proposed magnetic cilia studies. [23]Alfadhel Ahmed first produced cilia made of PDMS and Fe nanowires (Figure 3b).Pedro Ribeiro conducted similar research. [72]They mixed PDMS flexible material and NdFeB magnetic particles to make the magnetic cilia.Both of them can distinguish smaller forces, which can reach the μN level.In addition, magnetic cilia can also be used to distinguish surface roughness, [73] fruit maturity assessment, [74] and braille reading. [75]hile the cilia structure has higher sensitivity and a bigger detection limit than the film structure, the cilia are also more difficult to maintain and build in a vast area, and the consistency of the cilia has been confounding researchers all the time.After all, the cilia structures concentrate their efforts more on the high-sensitivity application, which assists people in restoring or improving their sense of tiny pressures.

Other Microstructures
The cilia structure and the film structure above are the common structures in the magnetoelastomer.In addition, a number of researchers have developed various convex microstructures based on the structure of films to fulfill the requirements of a variety of applications.In 2019, Ge et al. designed a pyramid-shaped extrusion on the magnetic film by a microstructure mold (Figure 3c).The pyramid-shaped design has the advantages of both cilia and film structure, and it improved the signal-to-noise ratio and broadened the pressure range.Their m-MEMS platform is an important milestone toward multifunctional, highly compliant human-machine interfaces. [46]In 2020, Ma et al. demonstrated a bioinspired self-powered tactile sensor by mimicking the shape and perceptual functionality of Merkel's disks.The integration of this magnetoelectric elastomer sensor into a robotic arm can help the robot distinguish between different types of subjects.The structure design of this self-powered sensor would make a significance for future intelligent robots and machine learning. [76]In addition, there are also some works using hemispherical and annular magnetoelastic structures. [42,77,78]Besides, using a 30 W ytterbium fiber laser to machine micro-holes, by doing this, the breathability of the magnetic film can be improved. [54]

Application
Due to a range of unique advantages associated with magnetic actuation, magnetoelastomers are currently widely applied in magnetic soft robots.[82][83] However, it's important to note that the primary focus of this review is on the utilization of magnetoelastomers in the context of human-machine interaction.Magnetoelastomers are still in the initial stage of human-machine interfaces.Since 2015, they have been continuously studied by researchers.The applications can be divided into contact applications and non-contact applications.Copyright 2019, The Authors, published by Wiley-VCH GmbH.

Contact Applications
The estimation of contact points, three-dimensional force magnitude, and direction are the main applications for contact magnetoelastomers.The magnetic sensors are necessary for contact applications, the magnetoelastomers are deformed by force, and the magnetic sensor senses the magnetic field change caused by the deformation, and then calculates the force parameters.Robot hands frequently employ touch applications for gripping and other tactile perception tasks.
In the beginning, the researchers mainly calculated the contact point and the normal force. [39] In addition, the researchers adopted the selfdecoupling and innovation support structure [50] to eliminate the force magnitude and direction.These studies provide a theoretical basis for contact human-machine interaction.For example, Alireza Mohammadi integrated multiple magnetic sensors with convex film structures into a manipulator and then used the manipulator to distinguish different objects. [78]Besides, Carvalho et al. developed a texture sensor with a cilia structure to distinguish fruit maturity assessment. [74]Yan et al. proposed a novel texture recognition method by designing a magnetic film and an attention-based long short-term memory (LSTM) model, which can efficiently recognize both braille characters and fabrics. [48,69,75]Magnetoelastomers are used in touch applications because of their deformation and tactile sensing capabilities.All these applications need to be equipped with magnetic sensors, which require many transmission signal lines, power lines, etc.Therefore, there are still a lot of limitations for HMI applications.Flexible magnetic sensors will be a good solution, but these researches are still in the beginning, and there is still a long way to go in the future (Figure 4).

Non-Contact Applications
Due to the non-contact characteristics of the magnetic field, direct contact is not required.Generally speaking, when it comes to non-physically touching, visual means are typically used for human-machine interfaces.Wearable magnetoelastomers provide the way for a new kind of non-contact human-machine interaction, which overcomes some of the limitations of visual approaches like sight occlusion.When magnetoelastomers are attached, people are able to create their own magnetic fields, making it easy to establish non-contact human-machine connections through the manipulation of magnetic materials.In 2018, Almansouri et al. proposed an imperceptible magnetic skin in which Ecoflex is used as the elastic matrix.The magnetic skin proposed has better stretchability, and it waived the sensation of its presence, when attached to the body, making it comfortable and imperceptible to wear. [27]After this, many magnetoelastomers attached to human beings applied to non-contact human-machine interaction.For example, Almansouri et al. presented a comprehensive assistive magnetic skin system that allows quadriplegics to move around individually and control their surroundings with ease. [54]Swanepoel et al. resent a versatile, robust, and facile system for tracking subcutaneous medical devices that entirely avoid X-ray and contrast agents.The system utilizes a magnetic skin to implement a permanent magnetic device tip. [60,61]Almansouri et al. used a magnetic skin system to track blinking and eye movements, which has many applications in automotive, gaming, sleep laboratories, controlling machines, and tracking our health. [40]Zhang et al. proposed a magnetic skin and applied it to gesture recognition and magnetic positioning, both perform well and provide a new way for future contactless human-machine interaction [33] (Figure 5).Lin et al, combined with the magnetoelastomers and the graphene oxide and carbon nanotubes, developed a flexible proximity sensor for touchless  [33] Copyright 2022, The Authors, published by Wiley-VCH GmbH.
tactile perception, including 3D position and motion tracking and human−machine interactions. [4]agnetoelastomers provide a novel path forward for noncontact HMI technology.Nevertheless, magnetoelastomers still require magnetic sensors in order to perceive, which will bring many limitations to the applications.In addition, the magnetic field generated by the magnetoelastomers without contact is relatively weak, which is easy to be affected by the external magnetic field (geomagnetic field), resulting in the performance degradation.Since it would be a smart move to develop a new type of wearable device integrated with flexible magnetic sensors and magnetoelastomers.There have been some similar reports at present, [46,84] but the researchers still need to make more efforts to widely apply the magnetoelastomers, even commercializing.

Flexible Magnetic Sensor
Section 2 explored in this review is the interaction between humans and computers using magnetoelastomers.Yet, the magnetic smart skin incorporates not only magnetoelastomers, but also flexible magnetic sensors.People can sense the magnetic field and interact with it through the flexible magnetic sensor.The first flexible magnetic sensor was provided by Parkin et al. who first grew GMR multilayers on Kapton, [85] and Mylar. [86]urther development was carried out by Chen et al. in 2008, who significantly enhanced the performance of GMR sensors on flexible substrates. [87]Years later, Melzer and colleagues used GMR multilayers on PDMS to develop the first stretchy magnetoelectronic sensors. [88]Then, in 2015, Melzer et al. presented wearable interactive devices that used Hall effect sensors, marking a significant step forward in the development of interactive technology. [89]Since then, a new field of human-machine interfaces based on magnetic electron skin has since been made possible.
, [84,105,106] presenting significant potential in biomedical applications.However, this paper focuses on the utilization of environmental magnetic fields or human-generated magnetic fields to facilitate communication between humans and computers.The advent of modern HMI can be traced back to the widespread use of flexible magnetic sensors used to showcase ideas such as virtual (VR) and augmented reality (AR). [107]In this article, we will review the most recent advancements in flexible magnetic sensors, including their substrates, sensor kinds, and potential applications.

Substrate
The substrate material is crucial to the development of the apparatus.Due to their high mechanical flexibility, low cost, and high-temperature stability, flexible polymers, rubbers, and metal foils are commonly applied as substrates in flexible magnetic sensors.Specific magnetoresistance sensors, such as spin valve giant magnetoresistance sensors, require thermal annealing to determine the magnetization direction of the pinning layer.As a result, the resist temperature of the substrate needs to be greater than the thermal annealing temperature.We will briefly review their physicochemical characteristics and potential uses (Table 1).

Polyimide
Marston Bogert first produced aromatic polyimides in 1908. [108]n 1955, high molecular weight aromatic polyimides were synthesized through a two-stage polycondensation of pyromellitic dianhydride with diamines. [109]Since then, interest in this class of polymers has been growing steadily.That is, polyimide (PI) exhibits excellent flexibility, extremely little creep, and high tensile strength, and it can be continuously maintained at temperatures up to 452°C. [110]And it holds up against ordinary organic solutions like ethanol and acetone, as well as mild acids and bases.The above advantages make PI compatible with the manufacturing processes of Micro-Electro-Mechanical Systems (MEMS), and it can be used as a substrate for manufacturing complex flexible magnetic sensors.Several groups have fabricated various flexible magnetic sensors on PI substrates and have demonstrated a small bending radius of 100 um and sensors on very thin (20 μm thick) PI substrates. [89,90,93,94,104]In addition, Kapton is one of the polyimide films, remains stable (in isolation) across a wide range of temperatures, used as a flexible substrate. [25,100]PI also has the property of being non-toxic, which makes it the most popular flexible substrate in wearable electronics. [95,111]

PET and PEN
Polyester or polyethylene terephthalate (PET) is a highperformance, crystal-clear thermoplastic made from ethylene glycol and dimethyl terephthalate (DMT) while polyethylene naphthalate (PEN) is a semi-aromatic, transparent polyester synthesized from naphthalene-2,6-dicarboxylic acid and ethylene glycol.PEN shares a chemical structure with PET, with the main distinction being that the PET molecular chain features a flexible benzene ring while the PEN molecular chain features a hard naphthalene ring.PEN's superior physical and mechanical qualities over PET are due to its naphthalene ring structure.However, PET and PEN are also two of the substrates that are typically used in flexible magnetic sensors.They have a transmittance of over 85% in the visible wavelength area, making them transparent to the eye, which is superior to polyimide. [89,90,93,94,104]esides, they have good stretchability and long-term performance.In 2015, the GMR sensor made by Melzer et al. maintained its performance under a 270% tensile strain and endured over 1000 cycles without fatigue, which is the best in all flexible magnetic sensors. [98,104]However, they are thermally much less stable than PI (<120°C) and can also be easily permeated by oxygen and water (water absorption of ≈0.14%).At present, serval researchers have proposed different types of magnetic sensors, such as AMR sensors, [103,112,113] GMR sensors. [91,92,98,106,114]and Hall sensors. [25,89,104]

PDMS
Polydimethylsiloxane (PDMS), also called dimethylpolysiloxane or dimethicone, is a kind of silicone and a polymeric organosilicon chemical.There are mature technologies and methods to endow PDMS with designer functionalities.Surface modification of PDMS can pattern its surface with adhesive and nonadhesive areas to selectively bond the active materials.Tailoring the composition of prepolymers before polymerization can alter their selfhealing, elasticity, modulus, and transparency properties. [115,116]ll of the PDMS properties, as well as the mature technologies to further enrich its functions render it as a good elastic substrate for stretchable electronics.
As a result, (PDMS) has been widely used as a substrate material in flexible/stretchable magnetic sensors due to its biocompatibility and intrinsic high stretchability (≈100-1100%).Many surface treatment procedures, such as oxygen plasma and ultraviolet radiation, have been developed to enhance the adhesion between the PDMS substrate and connected components.The PDMS substrate, on the other hand, has such weak adherence that it can be easily separated from the support.As a result, the flexible electronics on the PDMS substrate can be formed without the other redundant transfer process.][119][120] In particular, for TMR sensors, a higher annealing temperature is required, making PDMS substrates with relatively high roughness and lower resist temperature less suitable.

Silicon
Silicon is not a flexible substrate choice, but rather a very common rigid substrate.Yet, when the silicon substrate is reduced in thickness to a specific value, it is deemed flexible and is capable of keeping the magnetic sensor's good performance even in the bending state.In addition, devices on a silicon substrate can still be made into flexible wearable devices after being cut into small pieces and combined with flexible printed circuit boards.
In spite of the fact that atypical silicon substrates lack some of the flexibility and stretchability of other substrates, the roughness of silicon is superior to that of any other type of substrate and can ensure that sensors have superior performance.For example, Chen et al developed a different method: the backside of the silicon was directly etched in a deep trench etcher system to make a 14 um silicon substrate, [101] and Pérez et al. fabricated high-performance GMR sensorics on Si wafers, which are subse-quently thinned down to 100 μm or 50 μm to realize mechanically flexible sensing elements. [121]

Sensor Type
There are many types of magnetic sensors, including Hall sensors, anisotropic magnetoresistive (AMR) sensors, giant magnetoresistive (GMR) sensors, and tunnel magnetoresistive (TMR) sensors.Various types of magnetic sensors have distinct sensing methods and application ranges, and different sensors serve diverse purposes in various human-machine interfaces.Here, we summarized the properties of different types of magnetic sensors.The structure, sensitivity, MR ratio, and sensing range from various references are listed in the following table (Table 2).

Hall
A Hall effect sensor is a type of sensor that detects the presence and magnitude of a magnetic field via the Hall effect (Figure 6(i)).The Hall effect is a conduction phenomenon that is different for different charge carriers.Currently, Hall magnetic sensors are widely used in most of the applications and have achieved successful commercialization, mainly due to their well-developed Reproduced with permission. [89]opyright 2014, Wiley-VCH GmbH.Flexible AMR sensor: iii: Illustration of the AMR effort; iv: a) Fabrication process of the AMR sensor.b, Schematic of the device after fabrication and connection layout.Reproduced with permission. [103]Copyright 2018, Springer Nature.International license. [101]Copyright 2017, The Authors, published by Springer Nature.
and low-cost production. [122]The sensitivity, one of the most important factors for magnetic sensors, depends on the charge carrier mobility μ and the sheet carrier density n.Higher μ together with lower n results in larger sensitivity. [123]Figure 6(ii)) Several approaches have been taken to develop flexible Hall sensors, including the deposition of layered thin films (bismuth, [89] permalloy, [113] graphene, [25,104] and so on) on a flexible substrate.However, silicon-based Hall sensors perform well, while flexible Hall sensors do not.The low sensitivity of flexible Hall sensors still makes it difficult to conduct weak magnetic field detection.In recent work, a flexible and only 100 μm thick Hall sensor has been demonstrated using a thin bismuth layer.[89] Despite its good flexibility and a thickness of only 100 μm, the sensitivity of the device was about a factor of 40 lower compared to silicon-based Hall sensors, which is insufficient for many applications. In a4] Besides, graphene Hall sensors were combined with a deformable elastomer, and a flexible magnet to realize compliant soft tactile sensors. [25] Nevrtheless, even with advancements, Hall sensors still fall short when it comes to detecting magnetic fields that are too weak.

AMR
The AMR effect refers to the phenomenon that the resistance of anisotropic magnetic materials varies with the change of the angle between the magnetization and the current direction.The AMR effect is based on the anisotropic scattering of conductive electrons with uncompensated spins (Figure 6(iii)).The sensitivity of AMR sensors is improved, compared with the Hall sensors.Besides, the AMR sensor has low 1/f noise, which is more suitable for low-frequency magnetic detection.The direction of the magnetic field and the output response of an AMR sensor are correlated; thus, it is most commonly used in geomagnetic detection due to its high sensitivity to the direction of the magnetic field.In 2018, Canon Bermudez et al. developed a highly compliant e-skin compass based on the AMR effect.with capable of detecting geomagnetic fields (40-60 μT) with no loss of functionality even under bending to a radius of 150 μm. [90]Nevertheless, the magnetoresistance ratio of AMR sensors is ≈1-3%, and the performance of AMR sensors still lags behind other magnetoresistance sensors (Figure 6(iv)).If AMR sensors are to find widespread usage in human-machine interfaces in the future, they will need to fully exploit their advantages in easy manufacture and direction perception.

GMR
Giant magnetoresistance (GMR) is a quantum mechanical magnetoresistance effect seen in multilayers made of alternating ferromagnetic and non-magnetic conducting layers.Depending on whether adjacent ferromagnetic layers' magnetizations are aligned in parallel or antiparallel, there is a considerable difference in electrical resistance (Figure 6(v)).The resistance change of GMR sensors is substantially larger than that of AMR sensors, especially in the detection of weak magnetic fields.GMR sensors have been the most prevalent sensor type of magnetic smart skin in human-machine interfaces due to their great performance.
The GMR-based magneto-electronic skin is mainly constructed by depositing highly sensitive giant magnetoresistive (GMR) sensor elements on ultrathin foils.In 2015, Melzer et al. proposed a new flexible GMR sensor to offer a new sense for soft robotics, safety, and healthcare monitoring, consumer electronics, and electronic skin devices. [98]After this, many researchers proposed the flexible GMR sensor as a humanmachine interface. [46,91,92,98,106]And most of them tracked the movement of magnets.However, all previously mentioned are inplane sensitive GMR sensors, Makushko et al. demonstrated the first flexible magneto resistive spin valve switch operating with out-of-plane magnetic fields.This technology expands the functionality of magneto-receptive e-skins beyond basic proximity and orientation sensing (Figure 6(vi)). [92]n addition, GMR sensors can also be combined with other flexible films to create a multifunctional magnetically sensitive skin.Ge et al. combined the magnetic film and the flexible GMR sensor and proposed new a compliant magnetic microelectromechanical system (m-MEMS). [46]The system synergistically combines tactile and touchless interaction modes in a single sensor unit, and they showcased the usability of our bimodal e-skin in AR settings.
The GMR sensors have undergone significant development in the past 10 years and have a tendency to replace Hall sensors under many situations.However, the thickness, uniformity, and stability of the GMR magnetic layers are crucial for sensor performance.Non-uniform thin films or impurities within the thin films can lead to a decrease in performance.The nanoscale thickness of the thin films places high demands on the sputtering equipment.Additionally, spin valve GMR sensors require annealing to define the magnetic orientation of the pinned layer, typically at temperatures exceeding 150°C, which also places requirements on the substrate's temperature resistance. [37]Therefore, it cannot be widely used due to the limitations of the multilayers' deposition, the transition of the sensor, and the annealing temperature, although some researchers already constructed the first example of printed and stretchable giant magnetoresistive sensors. [93]The optimization of fabrication, however, will still be a primary focus for flexible GMR sensors.

TMR
Tunnel magnetoresistance (TMR) is a magnetoresistive effect that occurs in a magnetic tunnel junction (MTJ), which is a component consisting of two ferromagnets separated by a thin insulator.It is a dramatic change of the tunneling current in magnetic tunnel junctions when relative magnetizations of the two ferromagnetic layers change their alignment (Figure 6(vii)).The resistivity change of TMR is several times that of GMR sensors, and it also has higher sensitivity, which is more sensitive to the detection of weak magnetic signals in the application of humanmachine interaction.
However, reports of TMR in magnetic smart skins are scarce, and those researchers tend to focus solely on the production of flexible TMR devices rather than on any particular applications (Figure 6(viii)). [100,101]Besides, the structure of TMR is more intricate than that of GMR, the fabrication process is longer, the consistency is low, and it is difficult to achieve mass production.TMR devices typically require extremely thin oxide layers compared to GMR devices, and the preparation of these thin films demands higher thickness uniformity and surface flatness.Additionally, the manufacture of TMR devices necessitates more stringent process controls, including precise regulation of parameters such as temperature, atmosphere, and thin-film deposition rate, to ensure device performance stability and reliability. [37]In addition, the existing flexible substrates cannot sustain high temperatures, and the annealing of TMR devices typically requires a higher temperature, such as 350°C.The widespread use of TMR sensors has been hampered by all of these factors.Nonetheless, the TMR devices still have a superior magneto-resistance ratio and better sensitivity, which is beneficial to the development of magnetic-electronic skin.Therefore, TMR devices deserve extensive research in the future.

Application
Standard flexible magnetic sensors usually interact with the magnetic field source in the human-machine interfaces.The vectorial property of the magnetic field provides a new non-contact mode for human-machine interaction.These on-skin magnetic sensors allow control of the physical properties of objects and deliver some information in virtual reality relying on the interaction with ambient magnetic fields.The applications can be divided into three different parts: position interaction, direction interaction, and geomagnetic interaction.

Position Interaction
When it comes to human-machine interface, position interaction is more similar to a button that can facilitate the selection and typing of certain commands.Most proximity sensors are used for applications in position interaction that include the position relationship between a magnet and a magnetic sensor.The position interaction is carried out by the pre-instruction that corresponds to the approximation.This concept can be illustrated with a system consisting of a magnetic source, playing the part of a touchless button, which is approached by the user.In 2015, Melzer et al. demonstrated the first wearable interactive devices and applied it to position interaction. [89]They have introduced a technology platform that allows us to fabricate highly flexible magnetic field sensors relying on the Hall effect and fabricated an interactive pointing device by applying the flexible sensor to the finger.The relative position of the finger with respect to a permanent magnet is displayed in real-time by monitoring the sensor output.In addition, they further constructed highly sensitive GMR sensor elements on PET foils, the stretchable sensor is perfectly monitoring the pulsation of the diaphragm as the distance between the magnet and the compliant magnetoresistive foil changes during inflation and deflation.And it can be used for multifunctional medical implants. [98]fter that, many researchers on position interaction have been put forward, but these researches only focused on the output of magnetic sensor responding to a permanent magnet.Yet, when the magnetic sensor detects multiple permanent magnets with different polarity and remanence, the different instructions according to the difference of signals can be implemented.In 2021, Makushko et al demonstrated a novel magnetoreceptive platform for on-skin touchless interactive electronics based on flexible spin valve switches with sensitivity to out-of-plane magnetic fields, which is applied to the realization of magnetic ON/OFF interfaces based on the two magnets with different polarities (Figure 7). [92]Besides, the contact interactive devices can also be realized by attaching a layer of magnetic skin to flexible magnetic sensors. [46,92]n conclusion, GMR and Hall sensors are most commonly used in position interaction since there is no requirement for the sensitive direction of the sensors.Both GMR and Hall sensors are advantageous in that they can be mass-produced with little effort and provide a rapid response, making them ideal for use in situations where only the intensity of the magnetic field needs to be detected.

Direction Interaction
The position interaction of the previous section concentrated more on the detection of magnetic field magnitude, and the magnetic field is regarded as a scalar.Yet, the magnetic field is a vector, and it has direction.Consequently, if the magnetic sensor can determine the field's orientation, it can promote the humanmachine interfaces to a new level.
Angle changes can be detected using the AMR sensors, which have angular dependence.Besides, the AMR effect widely exists in ferromagnetic materials, and the AMR sensor can be composed of a simple sandwich structure (e.g., Ta/NiFe/Ta) in the form of rectangular strips.The simple structure makes it possible that the AMR sensor can be flexible without performance degradation.In 2016, Wang et al fabricated flexible AMR sensors on PET substrates, and the finalized AMR sensor showed a sensitivity of 42 T −1 which is close to that of the AMR on a rigid oxidized Si substrate. [112]The angle-sensitive AMR sensor has been used for pattern recognition of 10 μm thin magnetic strips formed by iron oxide magnetic ink.However, AMR ratio values of up to 30% have been discovered only at low temperatures, with just a small fraction persisting at room temperature.Besides, the AMR sensor has a limitation of the angular discrimination to just 180°.
The GMR sensors have the advantages of high sensitivity and a great range of magnetic ratios.Yet, the GMR sensors only have one sensitive axis, and it is impossible to sense the direction of the magnetic field with only one sensor.Therefore, it is usually necessary to use multiple GMR sensors to form a multi-directional sensitive sensing system.The local laser annealing or the transitionplacement assembly strategy is necessary for the direction interaction since many GMR sensors only have the same sensitive axis after one annealing.In 2018, Canon Bermudez et al. demonstrated a pick-and-place approach that combines thin film and transfer printing technologies to form eight spin-valve sensors in two perpendicular Wheatstone bridges.The sensors can discriminate between the x and y in-plane components of the magnetic field, and the virtual knob-turning and pressing functions can be achieved.However, the GMR sensors still only have two sensitive axes (x and y), and the 3D assembly methods [90] are difficult to carry out on flexible substrates (Figure 8).
As far, as direction interaction still has drawbacks of low magnetic ratio, trouble in fabrication, and limited in sensitive axis.However, the direction interaction occupies an important part, and it plays a dial-in human-machine interaction.Future applications in robotics, navigation, sports, and gaming could benefit from the more complicated programmable function of direction interaction compared to position interaction.

Geomagnetic Interaction
In the positioning and direction interactions, an extra external magnetic field source is required, which will limit the use of interactive devices.Also, a fully developed interactive device must be able to function in any environment and interact with its surroundings without restriction.In nature, there is a natural magnetic field source-geomagnetic field, and the application range of magnetic interactive devices can be greatly expanded if the flexible magnetic sensor can interact with the geomagnetic field.However, it is well known that geomagnetic fields are on the order of 0.4-0.6Oe and have different inclinations at different locations. [124]here are many types of magnetic sensors in geomagnetic field detection: for example, superconducting quantum interference devices, which require extremely low operational temperatures of liquid nitrogen in order to attain great sensitivity. [125]Fluxgate sensors [126] based on exciting coil/high permeability core have been investigated for many years to detect DC magnetic and geomagnetic fields.This widely used fluxgate sensor is relatively cheap and temperature-independent; however, its magnetic hysteresis, offset value under zero magnetic field, and large demagnetization factor restrict design considerations.
The high sensitivity in GMR and TMR sensors can be matched with the geomagnetic field, however, 3D detection is required in the geomagnetic field.The conventional 3D assembly process is difficult to achieve on flexible substrates.Besides, the GMR sensors need an extra bias, which is also difficult to achieve on flexible substrates to achieve a higher sensitivity.As far, only  [92] Copyright 2021, The Authors, published by Wiley-VCH GmbH.c) Movie snapshots demonstrating touchless manipulation of a virtual object using our on-skin 2D magnetic field sensor.The hand is turned with respect to the direction of the magnetic field lines of a permanent magnet.This motion of the hand is monitored, and the angular position is transformed into the setting of a virtual dial, in turn controlling the light intensity of a virtual bulb.Adapted with permission under a Creative Commons Attribution-NonCommercial License 4.0 (CC BY-NC). [90]Copyright 2018, The Authors, published by American Association for the Advancement of Science.
Ueberschär et al. fabricated a 2D sensor that yields a signal amplitude of 300 μV for a source voltage of 10 V in the Si substrate in 2015. [97] more viable alternative method is to use AMR sensors, which can easily integrate into flexible electronics and can be readily adjusted to detect the geomagnetic magnetic field.The AMR sensors can be matched with the geomagnetic field by conditioning with the barber pole method.[127] In 2018, Canon Bermudez et al. developed an e-skin compass by arranging AMR sensors geometrically conditioned in a Wheatstone bridge topology.They envisaged that this e-skin compass could enable humans to electronically emulate the magnetoactive sense, which some mammals possess naturally, allowing people to orientate with respect to Earth's magnetic field in any location (Figure 9).[103] AMR sensors, which can be easily integrated into flexible electronics and can be simply changed to monitor the geomagnetic magnetic field, are a more viable alternative method.These methods increase the sensitivity (giant magnetoimpedance) of an eskin compass and make it possible to detect the out-of-plane magnetic fields (Hall effect), both of which improve the device's functionality.

Prospect and Outlook
The use of magnetic smart skin is becoming more commonplace as we move deeper into the digital age.Magnetic smart skins will have a societal impact only when they go out of laboratories.The route from lab to end-user is thorny because the market for flexible sensors is complex.In this context, it is imperative to continue to develop HMI devices that are human-centered, that is to say, to seek the maximum functionality and performance of the device under the premise of biocompatibility.However, the development of magnetic smart skin is still in its infancy, and these devices have a number of issues with corresponding solutions as human-machine interfaces.We provide a prospect and outlook mainly focused on magnetic smart skins, sections 4.1,4.2,4.3 mainly focus on the discussion of both magnetoelastomers and flexible magnetic sensors, while sections 4.4 and 4.5 primarily emphasize flexible magnetic sensors.

Interface Stability
Interface stability refers to the interface between the human body and magnetic smart skin, as well as the interface between magnetoelastomers and flexible magnetic sensors.One of the most prominent mechanical challenges is the interfacial instability between dissimilar materials.Stress and/or strain concentration occurs at the above interfaces, leading to a major source of failure through delamination/detachment.The general principles in tackling interface instability are (1) improving interfacial adhesion and (2) avoiding abrupt softness/hardness difference.Specific methods vary in different scenarios, but the principles hold.
In addition, in the interface between the human body and the magnetic smart skin.All materials utilized in the creation of human-machine interfaces must be safe for human usage, making biocompatibility the very first and most crucial step.Although the PDMS and the flexible substrates have been proven to be nontoxic, these devices only have been verified in academic laboratories, and the long-term impact of these materials on the human body remains uncertain, limiting the commercial application in humans.Therefore, it is essential to do extensive research into new materials and elastic architectures.To improve breathability.For example, the micro-holes are realized using a 30 W ytterbium fiber laser in the magnetic skin to increase breathability. [54]Since the magnetic smart skin is linked to the human body, it must be able to tolerate the body's natural movements and shape changes without breaking.Improving fatigue resistance is important to sensor durability under cyclic loading and is particularly critical for magnetic smart skins.There has only been functional testing of magnetic smart skin on humans so far; reports on how well it holds up over time are not yet available.After all, durability is an important criterion for commercial devices.Therefore, Reproduced with permission. [103]Copyright 2018, Springer Nature.
fundamental studies on the failure mechanisms and failure criteria are important.

Performance
Performance is the next criterion for device functionality.For the magnetic skin, the stretchability and remanence are the most important properties, the durability and the using range of the magnetic skin can be through these two properties' improvements.For the flexible magnetic sensor, the balance and optimization of sensitivity, response time, limit of detection, measuring range, hysteresis, and other properties are constantly investigated.Performance stability is also essen-tial for deployable magnetic sensors because it ensures repeatable and reliable usage in changing environments and temperatures, especially for long-term use.Improving the environmental stability of sensor materials might be a straightforward approach.
Besides, the fabrication of the magnetic smart skin and electronics is still based on manual work, which increases the uncertainty of the device performance as well as the difficulty of mass production.Especially in GMR and TMR sensors, these two sensors have higher performance than other magnetic sensors, yet its fabrication process is difficult, and most of them can only be single-axis sensitive.As far, there is no flexible three-axis sensitive GMR or TMR sensor has been reported.Therefore, it is necessary to adapt the manufacturing method to mass-production platforms or investigate alternative processing methods, which can boost the market application of these low-cost soft devices.

Damage Insensitivity
Achieving and maintaining the functionality of magnetic smart skin, even in the presence of mechanical damage is the ultimate form of mechanical stability.To achieve this goal, mechanically tough and self-healing were developed.Traditional elastic materials, such as silicone rubbers commonly used in magnetoelastomers, tend to fail at notches due to their relatively low fracture toughness.To address this limitation, novel supramolecular elastomers with significantly enhanced fracture toughness have been synthesized. [128,129]Meanwhile, notable advancements have been made over the past decade in enhancing the toughness of stretchable hydrogels, [130,131] making them more stretchable and tough materials than some commercial elastomers.
In addition, self-healing materials have been engineered to autonomously repair mechanical damage, thereby enhancing the overall durability and longevity of the flexible electronics.While substantial progress has been achieved in the development of functional self-healing materials, including proximity sensors, [132] and humidity sensors, [133,134] magnetic field sensors, [135] self-healable sensors face many challenges for practical applications.Typically, the performance of self-healable materials and sensors lags behind that of their non-self-healing counterparts, and most devices that include self-healing materials simply demonstrate functionality.More effort should be directed toward understanding whether self-healing materials can bring value to durability and longevity at device and system levels.

Mass Production
The throughput of lab-scale production is extremely limited compared to industrial manufacturing.Lab-produced sensors often exhibit significant device-to-device variation, undermining the reproducibility and reliability necessary for commercial products.Additionally, during the transition to manufacturing, sensor and system performance should remain largely unaffected.
A general guideline for mass production dictates that manufacturing processes should be predominantly, if not entirely, automated, with a meticulous examination of process parameters.Printing offers a promising approach for integrating magnetic sensors onto flexible substrates.Printed electronics technology facilitates the creation of large-scale, low-cost magnetic sensors, instrumental in the development and manufacture of flexible sensors.It presents major cost and scalability advantages over conventional MEMS sensors.Research in printed magnetoelectronics has become highly captivating.[137][138][139] In particular, in Ref, [93] Ha et al. have demonstrated a stretchable highperformance magnetic field sensor relying on the GMR effect, which has a high sensitivity within 1 mT, and resists mechanical deformation of up to 16 μm of bending radii and 100% stretching.
New materials and manufacturing technologies typically require several years of development before commercialization.The semiconductor industry took several decades to evolve to its current scale.Fortunately, flexible magnetic sensors have the advantage of leveraging the existing semiconductor industry and other related sectors as steppingstones.We may not be far from their large-scale manufacturing.

Sensor Intelligence
Intelligence serves as a paramount distinguishing factor for future sensors compared to earlier sensor generations.Furthermore, intelligence represents a pivotal direction in the future development of flexible magnetic sensors.Human sensory systems exhibit remarkable efficiency while performing complex functions.Their effectiveness primarily stems from the organization and operation of the nervous system, which encompasses the brain, spinal cord, and peripheral nerves.Neuromorphic electronics have the potential to emulate certain aspects of the sophistication and efficiency seen in the nervous system for processing sensory information, making them a significant component of intelligent sensors.Conventional sensors are not inherently designed to efficiently implement neuromorphic sensory processing and computation.To bridge this gap, researchers have proposed and extensively explored various materials and device physics for neuromorphic devices. [140,141]Among these, neuromorphic devices based on spintronics, predominantly spintorque nano-oscillators, have already been demonstrated to be capable of implementing artificial neural networks.STNOs and TMR share a similar membrane structure and can be easily integrated on the same substrate in the future using the same microfabrication techniques, enabling more intelligent sensing systems, with features including analog computing and parallel storage and processing, in stark contrast to conventional digital processers.The STNO-based neuromorphic devices when integrated with magnetic sensors can preprocess sensory information in a delocalized manner, providing a promising route to edge computing. [140]However, much current work only simulates neural networks by extracting device parameters, instead of implementing physical demonstrations of array hardware. [142]Actual physical implementations face challenges in device yield and consistency, array integration, system robustness, etc.

Figure 1 .
Figure 1.Overview and conception of the review.

Figure 2 .
Figure 2. Three different types of magnetic field sources: a) Magnets.Reproduced under the terms of the CC-BY Creative Commons Attribution 4.0 International license.[41]Copyright 2016, The Authors, published by MDPI.b) Coils.Reprinted with permission from.[45]Copyright 2020 American Chemical Society c) Magnetic particles.Reproduced under the terms of the CC-BY Creative Commons Attribution 4.0 International license.[39]Copyright 2018, The Authors, published by MDPI.d,e) The characterization of the modulus of elasticity and remanence versus different content of the magnetic powder in PDMS and Ecoflex.Reproduced under the terms of the CC-BY Creative Commons Attribution 4.0 International license.[33]Copyright 2022, The Authors, published by Wiley-VCH GmbH.

Figure 4 .
Figure 4. Contact applications: a) The magnetic skin is sampled at 25 locations to a depth of 3 mm for a total of 25 classes.b) classification results for location 13, and c) all classification results grouped by class.Mean absolute error from linear regression grouped by location for d) x-position, e) y-position, and f) mean absolute error from k-NN regression for force.Reproduced under the terms of the CC-BY Creative Commons Attribution 4.0 International license.[52]Copyright 2019, The Authors, published by Wiley-VCH GmbH.

Figure 5 .
Figure 5. Non-contact applications: a) Gesture recognition system: the 3×3 sensor array circuit board collects the magnetic data under different gestures, and the manipulator makes the same gesture as the hand after processing.b) Magnetic sensors detect changes in the magnetic field caused by different gestures that can be classified through a neural network.c) Confusion matrix of different gestures.Reproduced under the terms of the CC-BY Creative Commons Attribution 4.0 International license.[33]Copyright 2022, The Authors, published by Wiley-VCH GmbH.

Figure 6 .
Figure 6.Flexible Hall sensor: i: Illustration of the Hall effect; ii: a) Single flexible Bismuth Hall sensor prepared on a commercial FPC.A confocal 3D microscopy image of the FPC head prior to the Bi deposition is shown as inset.b) Schematics of the Hall sensor cross-section taken along the geometrical center of the FPC.c) Integration of the flexible Hall sensor on a stator pole of a brushless electrical motor.Reproduced with permission.[89]Copyright 2014, Wiley-VCH GmbH.Flexible AMR sensor: iii: Illustration of the AMR effort; iv: a) Fabrication process of the AMR sensor.b, Schematic of the device after fabrication and connection layout.Reproduced with permission.[103]Copyright 2018, Springer Nature.Flexible GMR sensor: v: a) Illustration of the GMR sensor structure; b) Illustration of the GMR effort; vi: a) Schematic representation of the Co/Pd-based spin valve stack.b) Crosssectional TEM image of the spin valve deposited on a PEN flexible substrate.Reproduced under the terms of the CC-BY Creative Commons Attribution 4.0 International license. [92]Copyright 2021, The Authors, published by Wiley-VCH GmbH.Flexible TMR sensor: vii: a) Illustration of the TMR sensor structure; b) Illustration of the TMR effort; viii: a) Schematic of the MgO-barrier MTJ stacks.b) The bending behavior for a real MTJ devices when placed onto a Kapton film.Inset of b) shows the MTJ structures with a pillar size.Reproduced under the terms of the CC-BY Creative Commons Attribution 4.0 International license.[101]Copyright 2017, The Authors, published by Springer Nature.
Figure 6.Flexible Hall sensor: i: Illustration of the Hall effect; ii: a) Single flexible Bismuth Hall sensor prepared on a commercial FPC.A confocal 3D microscopy image of the FPC head prior to the Bi deposition is shown as inset.b) Schematics of the Hall sensor cross-section taken along the geometrical center of the FPC.c) Integration of the flexible Hall sensor on a stator pole of a brushless electrical motor.Reproduced with permission.[89]Copyright 2014, Wiley-VCH GmbH.Flexible AMR sensor: iii: Illustration of the AMR effort; iv: a) Fabrication process of the AMR sensor.b, Schematic of the device after fabrication and connection layout.Reproduced with permission.[103]Copyright 2018, Springer Nature.Flexible GMR sensor: v: a) Illustration of the GMR sensor structure; b) Illustration of the GMR effort; vi: a) Schematic representation of the Co/Pd-based spin valve stack.b) Crosssectional TEM image of the spin valve deposited on a PEN flexible substrate.Reproduced under the terms of the CC-BY Creative Commons Attribution 4.0 International license. [92]Copyright 2021, The Authors, published by Wiley-VCH GmbH.Flexible TMR sensor: vii: a) Illustration of the TMR sensor structure; b) Illustration of the TMR effort; viii: a) Schematic of the MgO-barrier MTJ stacks.b) The bending behavior for a real MTJ devices when placed onto a Kapton film.Inset of b) shows the MTJ structures with a pillar size.Reproduced under the terms of the CC-BY Creative Commons Attribution 4.0 International license.[101]Copyright 2017, The Authors, published by Springer Nature.

Figure 7 .
Figure 7. Position Interaction: a) Flexible SV switch affixed on a pointing finger and contacted in a 4-point configuration.b) Array of magnets used in the demonstration.The color map represents the distribution of the out-of-plane component of the magnetic field measured 1 cm above the magnets surface.c-f) Latching switch performance of the SV with 3 nm thick Cu spacer.c-e) Showing the interactive process by switching ON and OFF the navigation software on a virtual display.f) Top panel represents the temporal variation of the magnetic field, which is experienced by the SV switch upon its motion above the array of magnets.Bottom panel shows the corresponding voltage readout.g) Momentary switching performance of the SV with 2 nm thick Cu spacer.Timeline of the magnetic field applied to the SV (top panel) and corresponding signal readout (bottom panel).Reproduced under the terms of the CC-BY Creative Commons Attribution 4.0 International license.[92]Copyright 2021, The Authors, published by Wiley-VCH GmbH.

Figure 8 .
Figure 8. Direction Interaction: a) Skin-applied 2D sensor as on-skin electronics with directional perception.b,c) Movie snapshots demonstrating touchless manipulation of a virtual object using our on-skin 2D magnetic field sensor.The hand is turned with respect to the direction of the magnetic field lines of a permanent magnet.This motion of the hand is monitored, and the angular position is transformed into the setting of a virtual dial, in turn controlling the light intensity of a virtual bulb.Adapted with permission under a Creative Commons Attribution-NonCommercial License 4.0 (CC BY-NC).[90]Copyright 2018, The Authors, published by American Association for the Advancement of Science.

Figure 9 .
Figure 9. Geomagnetic Interaction: a) E-skin compass attached to the finger of a person.b) Time evolution of the output voltage of the e-skin compass when the person rotates back and forth from the magnetic north (N) to magnetic south (S) via west (W).c-e) Snapshots showing the instants when the person points to N, W and S. A compass rose dial with the cardinal points is overlaid on the snapshots to signal the corresponding orientations.Reproduced with permission.[103]Copyright 2018, Springer Nature.

Table 1 .
Comparison of different substrates.

Table 2 .
Comparison of different sensors.