Advances in magneto-ionic materials and perspectives for their application

The possibility of tuning magnetic material properties by ionic means is exciting both for basic science and, especially in view of the excellent energy efficiency and room temperature operation, for potential applications. In this perspective, we shortly introduce the functionality of magneto-ionic materials and focus on important recent advances in this field. We present a comparative overview of state-of-the-art magneto-ionic materials considering the achieved magnetoelectric voltage coefficients for magnetization and coercivity and the demonstrated time scales for magneto-ionic switching. Furthermore, the application perspectives of magneto-ionic materials in data storage and computing, magnetic actuation, and sensing are evaluated. Finally, we propose potential research directions to push this field forward and tackle the challenges related to future applications. reactions. This is promising for a drastic energy saving in comparison to conventional electric current-controlled magnetic devices in many possible applications.

by large spin polarized electric currents. For both cases, Joule heating and associated energy dissipation present a severe challenge. This problem continuously triggers numerous research activities on magnetoelectric (ME) materials 1-3 in which the magnetic material is controlled by the electric voltage instead of electric current, thereby reducing energy consumption.
Magneto-ionic control of magnetic materials is a novel approach in the field of magnetoelectricity. The magnetic material in this case is in contact with a solid or liquid electrolyte. By the application of an external voltage, ionic motion and electrochemical reactions are triggered which can reversibly affect the performance of the magnetic material (Fig. 1). As the field is rapidly evolving, very diverse material/electrolyte systems are studied, and several denominations and categorizations, such as redox-based, electrochemical, ion-exchange, or magneto-ionic control of magnetism, are in use. [4][5][6] In this perspective, we consider voltage-tunable magnetic materials systems in which the underlying mechanism is based on ionic motion and electrochemical reactions as magneto-ionic materials. The main advantage of magneto-ionic materials, in contrast to most other magnetoelectric materials, is that they can be operated at room temperature and at low voltage. Since chemical changes are involved, the non-volatile setting of a magnetic state by voltage is possible, requiring only a small current. This is in contrast to the volatile magnetoelectric effects achieved by capacitive charging via the electrochemical double layer and associated surface electronic structure changes in similar gating architectures. 7,8 Thus, magnetoionic materials offer a route toward a reversible manipulation of magnetic properties and are highly promising for ultralow power magnetic devices. FIG. 1. Magneto-ionic systems consist of a magnetic material in contact with a solid or liquid electrolyte. The magnetic properties, such as saturation magnetization and coercivity, are controlled by voltage-triggered ionic motion and electrochemical reactions. This is promising for a drastic energy saving in comparison to conventional electric current-controlled magnetic devices in many possible applications.
The first reports on electrochemical voltage control of magnetism in ultrathin metal films dealt with oxygen-based interfacial reduction/oxidation and associated changes in magnetic anisotropy and coercivity. 9-13 Since then, numerous magneto-ionic material systems consisting of magnetic metals and/or oxides in combination with a solid or liquid electrolyte have been investigated; an overview of these advances until 2018/19 can be found in recent reviews on magnetoelectric materials. [4][5][6]14 Within the past two years, significant progress has been achieved, such as magneto-ionic materials that function on the base of proton 15-17 or nitrogen ion 18 transport, the extension toward 3D magneto-ionic systems, 19,20 and novel magneto-ionic functionalities such as tunable exchange bias. 21,22 In this expanding field, most reports still deal with the exploration of fundamental phenomena. With the key advantages of energy efficiency and room temperature operation, however, targeted research toward specific applications is expected in the future.
In this perspective, we give an overview on important latest advances in magneto-ionic research, discuss the challenges and opportunities of magneto-ionic materials, and propose future research strategies.

II. RECENT ADVANCES IN MAGNETO-IONIC RESEARCH
In the field of magneto-ionic materials, significant advances have been made within the past two years, especially regarding room temperature operation, switching speed, reversibility, and the extension of magneto-ionic tunability to diverse magnetic phenomena and functional materials. To illustrate the broad range of magneto-ionic materials, Fig. 2 gives recent examples for voltage-tunable magnetization curves in ultrathin single metal films 15  In the following, we will focus on important advances from the past two years in light of the magnetic property, which is controlled by magneto-ionic means and with respect to application-relevant aspects, such as the switching speed and reversibility.

A. Magneto-ionic control of saturation magnetization
The voltage-control of the saturation magnetization (MS) by magneto-ionic mechanisms has been demonstrated in a large variety of systems. To give an overview on the attainable MS variations per defined voltage (V) change, the magnetoelectric (ME)-voltage coefficients, ΔMS/|ΔV|, 4 are plotted in Fig.   3 for selected magneto-ionic systems and as a function of the time between the two voltage-induced MS states. The data for very recent studies (from 2019 to 2020) are shown in red to highlight the current activities in the field. It should be noted that the time does not always represent the switching speed inherent to the material system, but sometimes also the time required for the magnetic measurement (which might be slower than the actual possible switching speed). Thus, the presentation in Fig. 3 gives the currently demonstrated time scales for magneto-ionic MS-control. Absolute values for the MS variation per volt [ Fig. 3(a)] are mainly reported for magneto-ionic systems based on ferrites and other transition metal oxides in thick film (≥100 nm) or in bulk form. 23-30 Several systems, with both the liquid or solid electrolyte, are directly derived from battery research here and measured with in situ magnetometry. 23,25,26 In some cases, the magnetically active electrodes are prepared by mixing transition metal oxide nanoparticles with binders and conductive nanoparticles. 23,24,26,27 This procedure optimizes the electrochemical operation, but at the same time, the magnetic moment per volume is diluted. Figure 3(a) shows that, so far, the largest ME-voltage coefficients for MS are in the range of several tens of emu cm −3 /volt. This seems small in comparison to conventional magnetic materials exhibiting high MS (e.g., CoFe alloy with MS up to 1950 emu cm −3 ); however, in the studied systems with small overall MS, the magneto-ionic effects are sufficient to modulate MS to a large extent. For instance, a 70% decrease in MS is reversibly achieved by Li intercalation in ZnFe2O4 for an applied voltage of just 1.25 V. 23 For most systems in Fig. 3(a), the time scales for the magneto-ionic measurements range from minutes to hours, which is very long with regard to applications. However, it must be acknowledged that most studies focused on the fundamental understanding of the mechanisms, targeted large tunable volumes, and utilized time-intensive measurements such as in situ superconducting quantum interference device (SQUID) magnetometers. 26,27 Recently, Ning et al. 30 reported non-volatile magneto-ionic control of MS in epitaxial SrCo0. 5Fe0.5O3−δ (SCFO) thin films via ionic liquid gating. Reversible and continuous control of the magnetization up to 100 emu cm −3 was demonstrated at room temperature, which was ascribed to voltage-induced changes in the oxygen stoichiometry. At negative voltage (−2 V), oxygen is extracted from the perovskite SCFO, which leads to the decrease in the magnetic moment (80% decrease in MS after 2 min). After prolonged gating, the nonmagnetic brownmillerite phase forms and thus ferromagnetic ordering is suppressed (MS ∼ 0 after 3 min). The ferromagnetic state can be recovered by applying a positive voltage (+2 V) at which oxygen is reinserted. In this way, reversible ON/OFF control of MS is obtained by alternating the polarity of the gating voltage. The switching speed of few minutes is remarkable, especially in view of the relatively large film thickness of up to 150 nm.
De Rojas et al. 18 proposed the use of nitrogen ions instead of oxygen ions to improve the magneto-ionic switching speed. They demonstrate a reversible ΔMS of 1 emu cm −3 upon the electrolytic gating of CoN (85 nm) thin films in propylene carbonate (PC) with solvated Na + species. To reach a ΔMS of 1 emu cm −3 , in this system, voltages of ±4 V and switching times around 4 min are required. A higher voltage of ±8 V and a longer time of 6 min are required to achieve the same effect in Co3O4 films of the same thickness. This difference between the magneto-ionic behavior of CoN and Co3O4 is discussed with regard to the lower activation energy for ion diffusion and the lower electronegativity of nitrogen compared to oxygen. In a further study, de Rojas et al. 28 revealed the importance of the gating geometry for the magneto-ionic efficiency for Co3O4. The insertion of a conductive buffer layer underneath the semiconducting and magneto-ionically active Co3O4 film (130 nm) to realize a bottom gate electrode yields faster switching and an increase in the ME-voltage coefficients from around 0.01 emu cm −3 V −1 to 0.11 emu cm −3 V −1 . 28 For the magneto-ionic control of Co3O4 28 and CoN, 18 in comparison to Li intercalation in spinel ferrite systems, 23 the ME-voltage coefficients are significantly smaller, but faster switching is demonstrated.
The relatively scarce data for absolute ME-voltage coefficients (with ΔMS in emu cm −3 ) in Fig. 3(a) are connected to measurement limitations for magneto-ionic materials, as they are often utilized in the ultrathin film or nanoporous form. One challenge is the difficulty to accurately determine the volume of these materials. A second challenge is that the very small magnetic moments, as in ultrathin films, are usually not measurable with conventional magnetometry, especially if in situ characterization requires further addition of the electrolyte and cell components. Therefore, for ultrathin films, surface-sensitive magnetic techniques based on the anomalous Hall effect (AHE) or the magneto-optical Kerr effect (MOKE) are utilized to probe relative changes in magnetization during magneto-ionic switching. 15,31,32 To give a more comprehensive overview, ME-voltage coefficients for the relative variation in MS over time are plotted in Fig. 3 Here, the great variety and extended activities regarding magneto-ionic material systems become obvious. Effective magneto-ionic control of MS can also be achieved in systems containing metallic ferromagnets, such as Fe, 33 Co, 34 FePt, 10 CoPt, 13 and Pd(Co). 16 In the FeOx/Fe system, the magneto-ionic switching relies on the reversible electrochemical transformation (surface oxidation/reduction) between weakly magnetic iron oxide and ferromagnetic iron metal in alkaline aqueous solution. 35 In this oxygen-ion based system, Duschek et al. achieved relative variations of MS up to 64% in continuous FeOx/Fe films (2 nm) 32 and almost ON/OFF switching of MS in FeOx/Fe nanoisland geometry 33 for a voltage change of just 1 V. The increased magneto-ionic effect in FeOx/Fe nanoislands, as compared to the thin film geometry, is attributed to the larger surface/volume ratio, which promotes the interfacial oxidation/reduction reactions. Recently, a continuous voltage-triggered ON switching of magnetism based on the reduction of paramagnetic β-FeOOH nanoplatelets to ferromagnetic Fe  16 demonstrated a fully reversible voltage-induced ON and OFF switching of magnetization for a voltage change of about 1 V. This strong magnetoionic effect is discussed with regard to electrochemical hydrogen sorption, which affects the magnetic coupling of the Co clusters via the Ruderman-Kittel-Kasuya-Yoshida-type interaction in the Pd matrix. In consequence, magneto-ionic switching between the ferromagnetic and the superparamagnetic state is achieved.
There have been remarkable advances also with regard to the voltage-control of magnetization in perovskite transition metal oxides by ionic mechanisms. 17,36,37 In superlattices comprised of alternating one unit cell of SrIrO3 and La0. 2Sr0.8MnO3, protons and oxygen ions can be transferred using ionic liquid gating, which triggers reversible phase transitions. 17 As a result, voltage-controlled ON/OFF switching of MS [ Fig. 2(c)] is demonstrated. Vasala et al. 36 studied the electrochemical (de)intercalation of fluoride ions into a two-layer La1. 3Sr1.7Mn2O7 system to tune the magnetization by exploiting the high sensitivity of the magnetic states to the Mn oxidation state and the distance between the perovskite building blocks. They achieve a 67% change in MS for low applied voltages (<1 V), which is reflected in a high MEvoltage coefficient [ Fig. 3(b)]. So far, the large magneto-ionic effects are restricted to a low temperature (e.g., 10 K) here.

B. Magneto-ionic control of magnetic anisotropy and coercivity
The control over magnetic anisotropy and coercivity (HC) is a key requirement for the design of magnetic materials in most types of applications. Figure 4 gives an overview on ME-voltage coefficients for HC achieved by magneto-ionic mechanisms, again as a function of the time passed between the settings of the two states. Most studies deal with ultrathin Co films exhibiting perpendicular magnetic anisotropy (PMA). 9, 11,12,15,38 In such films, the anisotropy and spin reorientation can be controlled by magneto-ionic mechanisms, resulting in an ON/OFFswitchable coercivity and remanence [ Fig. 2(a)]. Recently, large magnetoionic effects on anisotropy and HC have also been reported for Fe films with in-plane uniaxial anisotropy 39 and for typical hard magnetic materials such as L10 CoPt 13 and SmCo5. 40 This is a promising sign that magneto-ionic control can be transferred to a wide range of usable magnetic materials in the future. For the magneto-ionic control of anisotropy switching in ultrathin Co layers voltage-gated via GdOx, advances in switching speed and reversibility are notable. Tan et al. 15 demonstrated the benefits of protons over oxygen ions as the functional ions and achieved fast (100 ms) and highly reversible (2000 cycles) anisotropy switching at room temperature. In previously utilized GdOx/Co structures, which relied on oxygen ion migration, significant magneto-ionic effects required high temperatures or were restricted to the edges of the patterned electrodes. 9 For the proton-based mechanism, the authors suggested that protons, generated by atmospheric water splitting at the Au top electrode, are transported through GdOx to the Co interface. 15 For this mechanism, a sufficiently humid atmosphere is required. The magnetic anisotropy switching is ascribed to the high sensitivity of the interface anisotropy of Co to adsorbed H. In the same work, anisotropy switching is also observed for a Au/GdOx/Pd/Co/Pd structure, which is in this case explained by the hydrogenation of the Pd layer adjacent to the Co layer. Lee et al. 41 compared different proton conducting gate oxides and their effect on the switching speed for the magneto-ionic control of Co layers. The remanence ratio approaches zero within a time scale of >80 s for gadolinium doped ceria (GDC) and barium cerium yttrium zirconate (BZCY), while it takes less than 1 s for yttria-stabilized zirconia (YSZ). The Pt/Co/YSZ/Au structure is further optimized toward a Pd/Co/Pd/YSZ/Pt architecture in which Pt is used as top electrode to catalyze the electrochemical reaction and Pd serves as a hydrogen loading layer as well as a protection layer between the YSZ and the Co. Here, the fastest switching speed of <2 ms is achieved for a thickness of 10 nm for YSZ and with a gate voltage of +6 V. Huang et al. 42 recently showed that voltage-tunable PMA on the base of oxygen ion migration can also be achieved in Pt/Co/CoO/TiO2(TaOx) heterojunctions. In combination with the spin current generated from the Pt layer, this makes voltage control of the spin-orbit torque (SOT) induced perpendicular magnetization switching of Co possible.
Besides the substantial work on magneto-ionic control of ultrathin Co films with PMA, an extension to films with in-plane anisotropy has recently been reported. Zehner et al. 39  Recent advances are reported also for the magneto-ionic control of coercivity and magnetization in bulk-like materials. Navarro-Senent et al. 20 studied mesoporous Co-Pt/CoO microdisks (>300 nm) in PC with traces of NaOH. The voltage-driven oxygen migration promoted a partial reduction of CoO to Co and led to a strong decrease in coercivity (−85%) at −10 V, which could be partially recovered. In a subsequent study, Navarro-Senent et al. 44 investigated the behavior of nanoporous Co-Pt films covered by 10 nm AlOx or HfOx in the same electrolyte. Differences in the magneto-ionic behavior of uncoated Co-Pt films are ascribed to the oxygen acceptor behavior of HfOx and the oxygen donor behavior of AlOx.
For micrometer-sized SmCo5, Ye et al. 40 showed a huge modulation of coercivity (ΔHC of 1 T for ΔV < 1 V) by electrochemical hydrogenation in 1M KOH aqueous electrolyte. Even though very long timkes are required for the largest HC change, voltage-assisted magnetization reversal could be demonstrated within around 3 min. Further improvements in switching speed are proposed by the reduction of the particle size, the introduction of additional diffusion paths, or optimized device geometries.

III. APPLICATION PERSPECTIVES FOR MAGNETO-IONIC MATERIALS
In the following, application perspectives for magneto-ionic materials in the fields of memory devices, neuromorphic computing, spin-torque nanoantennas, sensors, and magnetic actuation are discussed.
Digital memory is a crucial part in our current digital era. The advent of internet of things, machine learning, and artificial-intelligence (AI) based applications, such as image recognition 50 or protein folding, 51 impose strict requirements on the processing speed and energy efficiency of data-centric tasks. Already today, the worldwide energy consumption of the data centers of all cloud service providers is about 200 TW/year. 52 Magnetic hard disk drives (HDD) remain a conventional way to store data in computers and large data centers, but the electric currents required for data writing are associated with the energy loss due to Joule heating. Also, charge based memories, such as flash and dynamic/static random-access memory (RAM), have high energy consumption and further device miniaturization becomes problematic mainly because of increasing leakage currents. 53 The use of magnetic elements for RAM seems appealing due to non-volatility and fabrication compatibility with CMOS. 54 Additionally, complex magnetic textures can be used for information storage. 55,56 In such magnetic memory concepts, electric currents are still used to control magnetization dynamics. The combination of magnetic memory concepts with energy-efficient magneto-ionic control could be a route to reduce the operational energies for next generation data storage media.
For HDD, control of coercivity or magnetization of the storage media by a voltage, instead of a magnetic field, is appealing to reduce the energy consumption. In high anisotropy magnetic thin films, which are required for data stability in high density magnetic data storage, a temporary reduction of coercivity during the writing process (magnetization switching) is desirable. So far, this is accomplished by heat-assisted magnetic recording, 57 which, however, involves the generation of heat with a laser. Here, provided that the switching speed is improved, magneto-ionic mechanisms to control coercivity may be a route to more energy-efficient voltage-assisted magnetic recording.
The device structure and nanoscale ion transportation mechanism in some magneto-ionic devices are comparable with those of resistive switching devices, which are used for the non-volatile resistive random access memory (RRAM). 58 It has been experimentally demonstrated that device resistance can be switched among multistates to achieve an ultra-high density/capacity storage. 59 In analogy, non-volatile ion transportation triggered by voltage could result in progressive changes in the magnetization value in magnetoionic materials. Thus, we propose that the stable multi-level intermediate states observed in many magneto-ionic materials could be extended into a multi-state magnetic storage concept.
Current-driven magnetic domain wall motion has raised hopes for new memory or logic devices, but so far, high threshold current densities and defect pinning effects pose challenges to the further development of this concept. 60 Magneto-ionic approaches have already shown that domain-wall propagation fields can be tuned through oxygen ion-based magneto-ionic control of ultrathin Co layers 61 and Fe films. 39 Offering the advantage of lower threshold current densities and topologically protection, 62 skyrmions are an alternative candidate to domain wall memory logic devices. One way to stabilize skyrmions is via the Dzyaloshinskii-Moriya interaction (DMI). By adjusting the strength of the DMI, it is possible to tune the domain wall chirality and the skyrmion's winding number, making it more suitable for practical applications. Interestingly, with regard to potential magneto-ionic control, DMI can be induced by oxygen chemisorption on a ferromagnetic surface. 63 A recent work demonstrated that the DMI strength can be (so far irreversibly) tuned by oxygen-ion migration at the ferromagnet/heavy metal (Co/Pt) interface. 31 Furthermore, gas phase experiments have shown that skyrmions can be induced in ultrathin Fe upon hydrogen exposure. 64 This indicates that modulation of skyrmions by fast proton-based magneto-ionic control may be feasible. 15 Fine tuning and reversibility of skyrmion parameters with high endurance, for example, through hydrogenation, could be favorable for the next generation spin-orbitronic devices.
In general, the limiting factors for future applicability in data technology might be, in the first place, the reversibility and switching speeds. In order to compete with the state-of-the-art technology, a stability over >10 15 cycles and switching speeds between 10 ms (HDD) and 10 ns [spin transfer torque (STT) RAM] are desirable. 65 The practical applications of AI require processing of large volumes of data, which, in conventional computing architecture (separate processing and storage units), is energy expensive and limited by von Neumann bottleneck and the memory wall. 66 Neuro-inspired computing chips that emulate the working principles of the biological brain are expected to perform better and more energy efficient. 67 Memristive elements are shown to emulate synaptic properties. 68 Since latest magneto-ionic approaches resemble the redox memristors in its operational principles, they may serve as an alternative building block for neuromorphic hardware. An artificial synapse and different synaptic functions based on the magneto-ionic setting of different magnetization states were already demonstrated in the Co thin film. 34 Magnonics is gaining a lot of popularity for wave-based computing 69 and microwave electronic 70 applications. It has been predicted that terahertz (THz) spin waves can be excited by domain wall motion induced by STT and SOT. 71 Large bandwidths of frequency modulation can be achieved, which is important for signal processing 72 and in telecommunication technologies. It was theoretically demonstrated in antiferromagnets that the wavelength of the emitted spin waves depends on material lattice constants. 73 One of the limitations to excite the spin waves is the high threshold current density in materials with strong exchange interactions. Here, the magneto-ionic mechanism such as ion intercalation or hydrogenation may be an approach to tune the lattice constants and the exchange interactions to obtain the desired wavelength.
Spin-torque nano-oscillators are magnetic tunnel junctions that emit microwave signals when a direct current induces sustained magnetization precession. A frequency tuning of 50 MHz has already been demonstrated in an electrically gated spin Hall nano-oscillator by engineering the electronic bands at the ferromagnetic/oxide interface. 74 However, an all ionic control of PMA would potentially lead to higher modulation of damping, 75 thereby directly influencing the auto-oscillation threshold current.
The opportunities to tune magnetic properties by a low voltage are also of interest for magnetic sensors. Modern magnetic field sensors rely on the giant magnetoresistance (GMR) or tunneling magnetoresistance (TMR) effect in functional layer stacks and are key to the information technology and numerous scientific and industrial measurement applications. Voltage control of GMR and TMR stacks via the electric field control of the magnetic anisotropy is proposed as a promising route to tuning and enhancing the sensitivity and the linear range of magnetic sensors. 76,77 With magneto-ionic approaches, ultralow voltage and non-volatile programming of magnetic sensors may be within reach, but this remains largely unexplored. One recent study on oxide/metal layers demonstrates a tunable magnetoresistance (MR) and even a voltage-controlled sign change in MR by magneto-ionic mechanisms. 78 A different type of potential sensor application relates to the interplay of electrochemistry and magnetism, which is inherent to magneto-ionic materials. The magnetic state not only depends on the applied voltage but also on the type of electrolyte species, pH, and hydration state, and thus, it can be regarded as a magnetic fingerprint to sense solution, solid-state, or atmospheric chemistry. Such a direct magneto-chemical sensor would have the advantage that functionalized magnetic particles, which are often utilized in magneto-chemical sensing platforms, 79 are not required. A first example in this direction is in situ characterization of (de)lithiation processes in LixCoO2 cathodes for Li ion batteries by utilizing in situ SQUID magnetometry. 80 Magnetic actuation is important in a plethora of modern technologies, including microfluidic chips, 81,82 (micro)robotics, 83 and MEMS. 84 Up to now, the magnetic actuation functionality requires an external magnetic field as the control parameter, which is usually provided by an electromagnet. This becomes problematic particularly when downscaling to the microscale, as here, high current densities are required to achieve sufficient magnetic field strength. 84 Switchable micromagnets by magneto-ionic control could present an energy-efficient alternative to standard micro-electromagnets, especially as the demand for high switching speed is less crucial than for memory devices. As an example, in some magnetophoretic devices, particle transport in a liquid above a magnetic film is realized by magnetic actuation in a time scale of seconds via the modulation of artificial magnetic domain patterns. 85 The required external magnetic field is so far realized by electromagnets. Direct magneto-ionic control of the magnetic domains 21 could potentially replace magnetic field control in such devices in a succinct way.
To achieve sufficient stray fields for actuation, a high energy product of the magnet is required. Thus, materials with high saturation magnetization, as a precondition for a high remanence, and sufficient HC will be favorable here. Furthermore, magneto-ionic materials exhibit the unique feature of different magnetic states (HC, remanence) being able to be set by an external voltage, and this opens the door to a voltage-programmable actuation functionality.

IV. POTENTIAL RESEARCH DIRECTIONS AND CONCLUDING REMARKS
While the technology readiness of this emerging topic is still at the fundamental research levels 1 and 2 (1-basic principles observed and 2technology concept formulated), the recent research works clearly show that magneto-ionic control can be extended to a broad range of material systems and magnetic functionalities. In the following, we propose potential research directions to push this field forward, especially in view of potential applications.
So far, magneto-ionic control is demonstrated for academic material systems and the material choice has primarily been conducted on the base of the targeted magneto-ionic mechanism. Now, with the basic knowledge on magneto-ionic principles, the research could extend toward magnetic materials, which exhibit optimized initial magnetic properties with regard to application. Here, research can rely on long-standing engineering approaches to design magnetic materials, including the optimization of the microstructure, texture, intergranular phases, or composition in alloys. For example, magneto-ionic control of bulk ferrites and rare earth hard magnets is so far restricted to isotropic materials. 29,40 An extension to textured materials could push forward the application potential for actuation, as the accessible energy product could be increased. Furthermore, the search for magnetoionic mechanisms to control antiferromagnetic layers, which are currently hot candidates for spintronic devices, seems promising. In this regard, Mustafa et al. 22 recently demonstrated voltage control of exchange bias by the intercalation of Li in antiferromagnetic Co3O4, placed adjacent to the ferromagnetic Co layer.
The energy efficiency of magneto-ionic systems is the most intriguing aspect for their application potential. Research works on magneto-ionic control already clearly demonstrate that a large variety of magnetic properties can be tuned by a low voltage. The electrochemical reactions involve electric currents, but these are orders of magnitude smaller than those required for electric current controlled magnetic devices. In addition, as the mechanism relies on chemical changes, the effects are often non-volatile and thus do not require a continuous power supply. For the FeOx/Fe system, magneto-ionic switching energies in the range of few femtojoules (for a 50 nm device) are projected, 39 which is comparable to the lowest reported switching energies for magnetic tunnel junction devices. 86 Considering that the switching speed is not yet optimized, this indicates that magneto-ionic mechanisms can yield an improvement in the energy efficiency of several orders of magnitude. Nevertheless, quantitative information on the energy consumption of magneto-ionic systems is still too scarce and should be provided more often in the future to highlight the competitiveness with alternative approaches in magnetic technologies.
The limitation of the switching speed, due to the required ionic motion, is considered as one of the main challenges for the application of magnetoionic materials. Consequently, future research should address this issue. For ultrathin Co films with PMA, which is the most-studied magneto-ionic material, the optimization of the device architecture, film thickness, and gate oxide has successfully increased the switching speed to the range of (sub)milliseconds 41 and, for small effects, also to the sub-nanosecond regime. 49 With this, these systems approach the required switching speeds for memory applications. Only for few systems besides Co, e.g., for the FeOx/Fe system, a switching speed in the range of seconds is reported, 39 which may suffice for actuation and sensor applications. Time scales of hours to days, which occur for many magneto-ionic systems (see Figs. 3 and 4), seem unacceptable from an application standpoint. In many cases, actually, the maximum possible switching speed may not be known due to measurement restrictions. Thus, a deeper understanding and an improvement of the key aspects for the switching speed is an important research strategy. In a first step, time-resolved magnetic measurements would offer access to the "true" switching speed. Second, interface-and defect-engineering could be pursued as a promising route to optimize the ion transport pathways and thus the kinetics, similar to the successfully applied concepts in the field of memristive systems. 87 The reversibility of magneto-ionic effects is a further important issue that defines the possible application areas. At first glance, magneto-ionic systems without a structural phase transformation (such as Li intercalation in spinel structure 24 ) or with purely interface effects (such as surface effects at ultrathin Co films 12,15 ) seem favorable here. In contrast, many magneto-ionic systems involve a phase transformation, which exponentiates tuning possibilities but may decrease reversibility due to strain-related fatigue or irreversible microstructure changes. In such systems, good reversibility is often found at the surface/interface regions or for nanoparticulate morphologies. 28,33,88,89 Thus, the minimization of the affected volume, e.g., via interface functionalization in nanostructured materials, seems to be a promising route to high reversibility. Furthermore, the unprecedented possibilities of magneto-ionic fine tuning of material properties by voltage at room temperature, even in the case of limited reversibility, should be acknowledged as a novel material design approach. Indeed, non-volatile magneto-ionic manipulation has already been proposed as the energy-efficient patterning method for artificial domains in exchange bias systems 21 or to set specific DMI values. 31 Magneto-ionic systems often involve liquid electrolytes. There are several advantages of liquid electrolytes over solid electrolytes, namely, the high electric fields that are accessible due to electrochemical double layer formation, the possibility to infiltrate 3D geometries, the easier characterization of solid/liquid in comparison to solid/solid interfaces, and the more straightforward understanding of mechanisms in the frame of conventional electrochemistry models. At the same time, the presence of a functional liquid is a critical point for implementation in most technologies. Therefore, research efforts should also target a transfer from solid/liquid systems to all-solid-state magneto-ionic systems wherever possible. Recent examples are the use of ionic gels, 90 hydrated oxides, 15 and devices that are derived from solid state battery architectures. 25,36 On a more futuristic level, solid/liquid magneto-ionic principles may become useful in brain-inspired computing concepts 91 or for magnetic actuation in microfluidic devices. 85 So far, research on magneto-ionic materials focused, to a large extent, on the basic science. Thus, in order to move from academic research to the application of magneto-ionic materials, the development of various laboratory demonstration prototypes is an essential future task.
In summary, this perspective gives an overview about important research advances in the field of magneto-ionic control of magnetism. These include the growing diversity of magneto-ionic materials systems and functionalities and the improvements toward high switching speeds. We discuss various application perspectives for magneto-ionic materials in memory, computing, actuation, and sensor technologies and suggest potential research directions. For instance, the quantification of switching energy and switching speeds and the development of magneto-ionic prototype systems seem important to highlight the energy saving potential and increase the competitiveness for future applications.

AUTHORS' CONTRIBUTIONS
M.N. and S.H. contributed equally to this work.

DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding author upon reasonable request.