Introduction

The advent of 2D van der Waals (vdW) materials1,2,3,4 has sparked a revolution in the field of magnetism, primarily due to the substantial influence of the dimensionality confinement and size effects upon spin textures and dynamics, e.g., quantum fluctuations. These magnetic 2D vdW materials (2D magnets) have garnered immense attention, driven by both their fundamental significance and facile integration into multilayer heterostructures. Ever since the experimental discovery of intrinsic 2D magnetic ordering in atomically thin layers5,6,7, the family of 2D magnets has expanded significantly over the years. Unlike conventional ultrathin film systems grown on substrates, e.g., CoFeB/MgO, these 2D magnets possess a naturally layered structure, high crystallinity, and appropriate coupling to transferred substrates of rational selection or design. Furthermore, the reduced coordination effects8, e.g., finite size effect, seen in ultrathin films are less pronounced in these layered magnets, thanks to their amenable interlayer interactions. Consequently, these 2D magnets serve as tailored testbeds for exploring the pure dimensional transition of magnetism from 3D to 2D. Additionally, these magnets exhibit high responsivity to external stimuli, such as gate voltage, molecule adsorption, and neighboring materials. The unique control of magnetism by electric means opens up possibilities for creating heterostructures and devices for applications in spintronics and memory technology.

While a lot of research is towards the discovery of intrinsic 2D magnets, 2D magnetic heterostructures aim at novel and exotic functionalities. The interaction between different layers in 2D magnetic heterostructures introduces new phenomena absent in the individual magnetic layer alone, which may expand their functionality for experiments and future applications. For example, we can tune the electronic band structures of 2D magnets via interfacial interactions in heterostructure construction. Artificial structures can be assembled to manipulate and enhance their magnetism through magnetic coupling and tuning. Furthermore, these 2D magnets can be harnessed to induce valley polarization and spin splitting in many other 2D materials, such as monolayer graphene9 and topological insulators10,11,12.

Based on these magnetic heterostructures, 2D spintronics is an emerging field to realize devices such as magnetic tunneling junctions (MTJ), spin field-effect transistors, and memristors12,13,14,15,16,17. 2D spintronics has unwrapped innovation and compelling opportunities that not only improve their performance but also diversify the functionality of electronic devices. For example, imprinted magnetic skyrmions can only exist at 2D interfaces18, which can be used as parallel information storage channels. When combined with electrical control, a switch of skyrmion on/off states can be realized19,20. Besides, 2D magnetic heterostructures for unconventional computing are an emerging field. One example is neuromorphic computing, which traditionally employs 3D bulk materials or nonlayered thin films, and thus the resulting device size is either difficult to scale down for ultrahigh-density integration or suffers from lattice mismatch problems. The emergence of 2D magnets offers a promising solution, as evidenced by the surge of reported 2D heterostructures functioning as neuromorphic computing devices21,22, as well as using skyrmions23,24,25. In parallel, at the quantum computing frontiers, concepts such as 2D magnetic heterostructures with superconductors towards topological superconductivity26,27 and skyrmion qubits28,29 with scalability, offer easy electric control and new functionality. In this review, we first present the large family of existing 2D magnets, including ferromagnet, antiferromagnet, and multiferroics, and then discuss the multifunctionality in 2D magnetic heterostructures. Based on these heterostructures, 2D devices for spintronics will be examined, with a target for applications in high-performance and high-density magnetic memories, as well as neuromorphic and quantum computing.

Library of 2D magnets

The hallmark of 2D magnetism is the existence of an ordered arrangement of magnetic moments over macroscopic length scales at any finite temperature, with a spontaneous breaking of time-reversal symmetry. Since the experimental discovery of intrinsic magnetic ordering in 2D materials, a rich collection of 2D magnets covering a wide spectrum of magnetic properties has been reported5,6,7,30,31,32,33,34,35,36,37,38,39,40,41, as shown in Fig. 1. Many of these materials are semiconductors with band gaps covering the near-infrared to the ultraviolet spectral range, whereas a few are metallic.

Fig. 1: Illustration of 2D magnetic materials, showing the critical temperature and electrical properties.
figure 1

Arrangement is in accordance with their magnetism, conductivity, and critical temperature. Ferromagnetic and antiferromagnetic materials are characterized by Curie and Néel temperatures, respectively. They are divided into three dashed spheres based on conductivity: smallest (insulating), medium (semiconducting), and large (metallic). Multiferroics are highlighted in red color.

Full size image

Ferromagnetic (FM) order exists in both layered transition-metal halides and chalcogenides when the exchange interactions favor both intralayer and interlayer parallel spin alignments. The Curie temperature TC is known as the temperature above which the material undergoes a magnetic transition from a ferromagnet to a paramagnet. It is relatively easy to control and manipulate the magnetization direction in FM materials compared to antiferromagnets, making them suitable for various applications in magnetic memory and spintronics.

Antiferromagnetic (AFM) materials have a rich variety of magnetic orders. They can be broadly divided into two types: those with intralayer FM order and interlayer AFM order, and others with intralayer AFM order. Antiferromagnets are magnetically ordered, yet with zero magnetization, making them immune to external field perturbations. Their AFM order is established by the exchange interactions between the spins, which produces an exchange field in the order of 102 to 103 T (one or two orders of magnitudes larger than in the FM case) and gives rise to a THz spin precession frequency. Thus, devices based on AFM materials usually exhibit intrinsic stability and can potentially enable high-speed operation, making the proximity effect between AFM and other materials extremely promising. In the following, we discuss some magnetic materials under extensive studies, with physical properties promising for device design.

CrI3 is an example of 2D magnets. Atomically thin CrI3 flakes were successfully prepared through mechanical exfoliation6. Subsequently, monolayer FM properties were demonstrated using low-temperature microscopic magneto-optical Kerr effect (MOKE) measurement techniques, revealing a Curie temperature of 45K. Notably, there is strong evidence indicating layer-dependent magnetic phases: monolayers exhibit ferromagnetism, bilayers display antiferromagnetism, and trilayer/bulk configurations revert to ferromagnetism. Studies also underscore the challenge of comprehending the magnetic properties of thin layers of CrI3, as FM and AFM orders can coexist in the same flake40,42. But this complexity also presents novel opportunities to engineer magnetic states and achieve innovative functionalities.

The intrinsic ferromagnetism in Cr2Ge2Te6 (CGT) layers was reported around the same time as the study of CrI37, and it was also revealed by MOKE microscopy under a small stabilizing magnetic field of 0.075 T. CGT is an FM material with a band gap around 0.8 eV43. Its electrical property was characterized by a two-terminal resistivity of around MΩ at room temperature44, showing the electrically resistive nature. Additionally, CGT thin layers exhibit a thickness-dependent TC, from about 30 K for a bilayer to 68 K at its bulk limit. This thickness dependence of TC exists also in as-grown thin magnetic films, which show more pronounced dependence45.

The two materials mentioned above have relatively low TC, with respect to room-temperature memory devices. To enable the application of 2D magnets in spintronic devices, the emphasis is on developing 2D materials with stable magnetic properties at room temperature. Among all 2D magnets, Fe-based materials usually have higher TC, e.g., Fe3GeTe2 (FGeT). Its itinerant ferromagnetism was observed to persist down to the monolayer30, due to its strong perpendicular magnetic anisotropy (PMA) with a uniaxial magnetocrystalline anisotropy constant Ku = 1.46 × 107 erg/cm346. This anisotropy energy is about two or three orders of magnitude higher than those of CrI3 (Ku = 4.3 × 104 erg/cm3 at 50 K)47 and CGT (Ku = 1 × 105 erg/cm3)48. In bulk FGeT, the TC is 200 − 220 K, depending on Fe deficiency. Although this value is diminished in atomically thin FGeT flakes, it can be significantly elevated to reach room temperature by employing liquid gating49, showing the electrically tunable nature of these 2D magnetic materials.

Another example is VSe2, with monolayer grown by molecular beam epitaxy (MBE) method. This material has been shown to maintain its FM ordering at 330 K31. Exceptionally, a recently discovered material Fe3GaTe2 (FGaT), with a similar crystal structure to FGeT, exhibits itself as an above-room temperature 2D metallic ferromagnet32. It has been demonstrated to offer a record-high TC ( ~ 350–380 K) as well as a robust PMA. These outstanding magnetic properties render FGaT a widely favored material for practical magnetoelectronic and spintronic applications.

2D magnets can also be made free-standing to enable mechanical degrees of freedom, dynamical motions, and oscillations, thus opening new opportunities for studying mechanical control and phononic coupling with designed structures of 2D magnetic materials, especially on the 2D magnetic nanoelectromechanical systems (NEMS) platform. 2D AFM CrI3 NEMS resonators have been employed to probe the magnetostriction effects50. 2D AFM FePS3, MnPS3, and NiPS3 membrane NEMS resonators have also been exploited to measure varying-temperature-induced phase transition in these 2D magnets51. All these 2D magnetic NEMS resonators have also demonstrated excellent tunability of their resonance frequencies via electrostatic gating effects50,51. Furthermore, toward coherent mediation and dynamic manipulation, phononic crystals built on 2D materials and heterostructures enable phonon band-selectivity, cavities, and waveguides52, which are being actively studied to facilitate coupling phonons to spin dynamics, propagating spin waves (magnons), and magnetic vortices (skyrmions).

Beyond magnetic properties, these 2D magnets also provide a unique platform for the realization of topological magnetic materials, enabling the study of the interplay between magnetism and topology. MnBi2Te4 and its derivative compounds have received focused attention recently for their inherent magnetic order and the rich, robust, and tunable topological phases12. MnBi2Te4 has an AFM interlayer coupling, while an FM order exists in every single layer, with an out-of-plane easy magnetic axis. The interplay between the magnetic structure and the topological nontrivial bands endows the materials with rich topological phases, e.g., quantized anomalous Hall effect above 20 K33,53, which has a higher temperature than magnetically doped BST samples54. In odd-layer flakes, a quantum anomalous Hall effect was observed with quantized plateaus of Hall resistivity as well as diminishing longitudinal resistance55, absence of any applied magnetic fields. This is extremely promising for low-power devices with dissipationless edge states. These materials and their heterostructures provide a neat, simple, and versatile way to fuse topological bands with magnetic order, which are promising to help elevate the quantum anomalous Hall effect towards liquid nitrogen temperature. Possession of topology contributes to improving the performance of electronic components, by enabling high fault tolerance with topological protection, which is a key factor desired in quantum computing.

Additionally, 2D multiferroic materials are actively studied for their cross-coupling effects. This class of materials exhibits multiple ferroic properties simultaneously, including ferromagnetism, ferroelectricity, and ferroelasticity. Multiferroics hold great promise for the development of innovative, multifunctional devices, with enhanced device performance and ultra-scaled sizes in vdW cases. The discovery of type-II multiferroic order in a single atomic/molecular layer of the transition-metal-based layered material NiI2 was reported34, which exhibits an inversion-symmetry-breaking magnetic order and directly induces a ferroelectric (FE) polarization. Later, intriguing in-plane electrical and magnetic anisotropy in layered multiferroic CuCrP2S635 has generated much excitement owing to the coexistence of antiferroelectricity and antiferromagnetism, along with strong polarization-magnetization coupling. These multiferroics enable magnetization affected by electric fields, and vice versa. This functionality has motivated fundamental and applied research efforts in applying multiferroics for voltage-controlled magnetic anisotropy, toward low-power switching devices.

Overall, the current main thrusts for 2D magnets search are (1) room-temperature atomically thin magnets, for high-density and high-performance memory; (2) FM or AFM semiconductors with a tunable band gap and stability, which would be an ideal platform to investigate the magnetoelectric coupling between electrons and spin; (3) exotic topological properties, like searching for high-temperature quantum anomalous Hall insulators; (4) multiphysics coupling effects in engineerable device platforms, such as investigating electromechanical tuning of 2D magnetic membranes, and interactions between phonons and magnons; (5) large-scale thin films for device integration. Although exfoliated 2D films offer high-quality device layers, the limitation on the lateral size normally to hundreds of micrometers still hinders their applications at the level of integration and manufacturing. Efforts are currently ongoing to obtain high-quality wafer-scale 2D magnetic heterostructures.

2D magnetic heterostructures and their interface

2D magnetic heterostructure is a material system composed of two or more distinct layers of atomically thin magnetic materials stacked together. These layers can have different magnetic properties or orientations. The use of 2D magnetic heterostructures is mainly for two purposes. One is to expand the functionality of these structures for both experimental investigations and future practical applications. The other is because the proximity effects within heterostructures give rise to valuable phenomena that enhance the magnetic properties. When combined with magnetism, the interface has even more intriguing physics and phenomena as shown in Fig. 2, compared to the charge-based system or individual magnet.

Fig. 2
figure 2

Schematic mechanisms of the 2D interface in magnetic heterostructures, including orbital hybridization, exchange coupling, lattice perturbation, strain effect, SOC proximity, and magnetoelectric effect.

Full size image

Firstly, the orbital hybridization at the interface can not only induce the exchange interaction at the interface but also modify the electronic and magnetic properties in adjacent layers by impacting their band structures and orbital characters. Orbital hybridization is a new degree of freedom for controlling interfacial Dzyaloshinskii-Moriya interaction (DMI), a short-range anti-symmetric exchange interaction56. Taking CGT/FGeT heterostructure for example, the topological Hall effect arising from magnetic skyrmions has been observed, which is due to interfacial DMI18.

Another proximity effect is exchange coupling, which refers to the interaction between magnetic moments of different layers within the heterostructure. It’s crucial for applications in magnetic memory and spintronic devices, like spin valves, MTJs, and magnetic sensors based on giant magnetoresistance (GMR) or tunneling magnetoresistance (TMR). As reported, large exchange bias was observed in an FGeT/antiferromagnet interface57, which is tunable by magnetic field cooling. Antiferromagnet, with zero magnetization, shows robustness against external perturbations and THz fast spin procession. These characteristics make coupling to AFM materials extremely promising.

Following is lattice perturbation, which distorts the crystal lattice structure at the interface, thus tuning the magnetic behaviors. Examples include applying electric voltages and vertical stacking to form the Moiré lattices58,59,60. This can lead to either the control of magnetization or the emergence of new exotic effects like Moiré skyrmions61.

Different from the above mechanisms, the strain effect arises from mechanical deformation, leading to a change in electronic band structures. The methods to induce strain include lattice mismatch, utilization of flexible, patterned, or piezoelectric/FE substrates, electrostatic/capacitive gating62,63, electromechanical actuators64, as well as the atomic force microscopy tip, or even the presence of bubbles at the interface during sample assembly, or engineered bulging and micro-bubble-blowing techniques65.

Spin-orbit coupling (SOC) refers to the interaction between the spin and orbital angular momentum of electrons across different layers within the heterostructure. This can play an important role when the 2D magnet is interfaced with a heavy metal element. An example of this is the emergence of skymions in WTe2/FGeT heterostructure66 through SOC coupling. This SOC is also crucial to realize spin-orbit torque (SOT) or spin-transfer torque (STT) in magnetic random-access memory (MRAM).

The last yet significant effect is the magnetoelectric effect, which occurs when an FM layer interfaces with an FE or multiferroic layer. The magnetism of the FM material can be switched by the electric polarization in the FE layer, and vice versa. FE control of magnetism is particularly interesting due to its energy-efficient and non-volatile nature. One example is to use a thin FE polymer to open and close the CGT magnetic loop67.

An alternative promising approach to generating a substantial magnetoelectric effect involves the stacking 2D FM layer with recently emerging fluorite68,69,70 and wurtzite FE71,72,73 materials. These CMOS-compatible FE materials, such as hafnium-dioxide (HfO2), zinc-magnesium-oxide (ZnMgO), or aluminum-scandium-nitride (AlScN)74, offer a unique advantage over conventional perovskites due to their remarkable scalability, while providing very large polarization and coercive fields75. Moreover, fluoride FEs with sub-nm thickness76,77 and wurtzite FEs with sub-10 nm thickness78 have been demonstrated, and their integration with 2D materials to create novel memory devices was studied79. In comparison to 2D FEs80,81 as shown in Table 1, these 3D materials offer one to two orders of magnitude higher polarization. Further decreasing thickness to below 10 nm in these 3D materials would be very challenging, where the size effect would weaken or even eliminate the ferroelectricity82. In 2D layered FEs, the dependence of ferroelectricity on thickness is quite diverse in different materials. For instance, theoretical analysis suggests that ferroelectricity exhibits an odd-even effect in some 2D materials, e.g., SnSe83, whereas the piezoresponse in CuInP2S6 (CIPS) exhibits an increasing tendency as thickness increases84. The polarizations in 2D FEs exhibit comparable values with those of the bulk, thus suggesting the great potential of 2D FEs in high-performance nanoscale devices. The high-quality and seamless 2D interfaces offer a platform for introducing new functionalities into 2D devices, thereby enabling the exploration of novel physics and effects. Of particular interest is the pursuit of ultra-fast, ultra-compact, and energy-efficient spintronics through precise interface engineering. By leveraging the interface, it becomes possible to amalgamate THz spin procession in AFM materials, non-volatile control via 2D FEs, and scalable configurations into a single device. This development holds promise for advancing the field of spintronics in the future.

Table 1 Comparing FE properties between different classes of 2D and thin materials
Full size table

2D Spintronics

Device for magnetic memory

Introducing magnetic elements into memory devices provides substantial advantages, particularly in achieving non-volatile memory functionality. The utilization of 2D magnetic heterostructures, comprised of multiple layers of 2D materials with unique properties, takes these advantages even further. It enhances energy efficiency, boosts data storage density, and enriches the range of functionalities achievable within a single memory device, as shown in Fig. 3.

Fig. 3: Schematic of 2D magnetic heterostructures for magnetic memory.
figure 3

Examples include magnetic tunneling junction (MTJ), spin-tunnel field-effect transistor (spin-TFET), spin-orbit torque (SOT) devices, and Josephson junction.

Full size image

MTJ is a fundamental building block in modern memory devices. It consists of two FM layers separated by a thin insulating barrier. The core principle governing MTJ operation is the TMR effect, wherein the electrical resistance of the junction changes based on the relative alignment of magnetization in the two FM layers. In contrast to traditional bulk ferromagnets such as Fe3O4 and Co, employing atomically thin layered magnets for MTJs offers the potential to significantly enhance their TMR, since TMR heavily depends on interface quality. Additionally, this approach relaxes the stringent lattice-matching requirements associated with epitaxy growth and enables high-quality integration of dissimilar materials with atomically sharp interfaces.

Among the existing 2D magnets, CrI3 stands out due to its interlayer AFM ordering for bilayers and relatively electrically insulating properties, making it a desirable choice as a tunnel barrier in MTJs. One spin-filter MTJ design involves the insulating CrI3 thin layer as a tunnel barrier sandwiched between graphite contacts85. This device demonstrated remarkable TMR values of 530%, 3200%, and 19,000% for bilayer, trilayer, and four-layer CrI3 structures, respectively. The highest value of the four-layer structure greatly exceeds typically achieved in conventional bulk material MTJs86,87. While MTJs based on CrI3 exhibit promising properties in TMR, it is important to acknowledge their operational constraints—cryogenic low temperature for operation, instability in the air, and need for external magnetic fields. In response, ongoing research focuses on advancing MTJs by employing materials suitable for room-temperature operation. A notable candidate is metallic FGaT, which boasts a high TC of 380 K, as well as an operation temperature above room temperature. This characteristic overcomes the limitation, greatly facilitating the realization of practical spintronic devices. A room-temperature MTJ was based on a WSe2 spacer layer embedded between two FGaT electrodes with intrinsic above-room-temperature ferromagnetism88. This device showed a significant TMR of 85% at room temperature.

Spin field-effect transistor (Spin-FET) is a new type of spintronics device for non-volatile memory. This device utilizes the spin, rather than charge, in the flow of electrical current, which could potentially offer non-volatile data storage and improved performance compared with traditional FETs. The application of a magnetic tunnel barrier in MTJ can be extended to Spin-FETs, leading to a novel device—the spin tunneling FET, e.g., dual-gated graphene/CrI3/graphene tunnel junctions89, with a similar structure as spin-filter MTJs. This device achieves an impressive high-to-low conductance ratio of nearly 400% by electrical modulation of magnetization configurations. However, it shares a common constraint with CrI3-based MTJs, requiring low operational temperatures and magnetic fields. Therefore, 2D magnets with high critical temperatures hold the key to advancing research, since they ensure the stability of 2D magnetic devices at room temperature, enhance their practical applications, and facilitate their seamless integration into existing technologies.

Utilizing SOC, SOT devices are engineered to manipulate the magnetization of a magnetic layer by a spin-polarized current. Compared to traditional STT devices, SOT devices are anticipated to offer superior performance like easy stacking, operation in sub-ns writing and reading speed. Typically, a SOT device comprises two key components: an FM layer and a non-magnetic heavy metal (HM) layer. SOT efficiency heavily relies on the interface quality. Thus, 2D magnets are promising building blocks for SOT devices90,91, like CGT, CrI3, and FGeT with intrinsic PMA. Even in monolayer form, these materials maintain their PMA, making them highly responsive to external stimuli. In the initial designs of 2D magnetic SOT devices, structures like Pt/FGeT92 and Ta/CGT93 were featured, showing significant improvements in SOT switching efficiency with the incorporation of 2D FM materials. Creating all-vdW heterostructures can further enhance the switching efficiency. For instance, a FGeT/WSe2 heterostructure was realized90, achieving a remarkably low switching current density of 3.90 × 106 A/cm2 at 150 K, which is comparable to, or even better than, those of conventional thin films and their systems94,95. Recent research efforts have been directed toward achieving reliable vdW SOT operation at room temperature. High TC FGaT was incorporated to achieve room temperature operation96, based on an FGaT/Pt bilayer structure. It requires a switching current density of 1.3 × 107 A/cm2. Another notable approach involves a wafer-scale FGeT/Bi2Se3 layered heterostructure97, where the topological insulator Bi2Se3 plays a crucial role in increasing the TC of FGeT to room temperature through exchange coupling. This approach yielded a high damping-like SOT efficiency of ~2.69 and a notably low switching current of 2.2 × 106 A/cm2 at room temperature90.

Apart from conventional memory, Josephson junction (JJ) combines the quantum phenomena with memory device technology, opening the door to the development of superconducting random-access memories (RAMs)98. JJs are typically composed of two superconductors (SCs) separated by a thin insulating or non-superconducting barrier. JJ based on 2D ferromagnet/SC heterostructure presents a unique opportunity to explore the interaction between superconductivity and ferromagnetism. Compared to ultrathin magnetic films, the stronger surface and proximity effects of 2D materials can lead to enhanced interplay between these two layers99,100. Additionally, some 2D materials with topological insulator behavior, featuring impurity-resistant edge states, offer advantages when studying the coexistence of superconductivity and ferromagnetism101,102. For the device example, one approach involves constructing a JJ by introducing a few-layer FM insulator, CGT, between two layers of SCs NbSe2103. This JJ demonstrates a hysteresis behavior in the critical current, which is induced by the remanence of the magnetic barrier. Following this, another lateral JJ has emerged, comprised of an FGeT, laterally interconnected between two layered spin-singlet SCs NbSe2104. This SC/ferromagnet/SC heterostructure successfully sustains skin Josephson supercurrents with a remarkable long-range reach (depth) over 300 nm.

2D spintronics for neuromorphic computing

Neuromorphic computing is an emerging field of artificial intelligence (AI) that seeks to design computer systems inspired by the human brain’s neural architecture. Owing to the atomic thickness, dangling-bond-free surfaces, and high mechanical robustness, 2D materials and heterostructures are extensively investigated for neuromorphic computing devices, showing great promise for high-performance artificial neurons and synapses105,106,107,108,109. The integration of magnetic elements within 2D materials to form 2D magnetic heterostructures takes this field a step further. These magnetic elements provide non-volatility, longevity, and high-density information storage capabilities. Furthermore, 2D magnetic heterostructures have the capability to generate skyrmions18,66, which are topological spin structures with unique topology towards more compact, energy-efficient, fault-tolerant, and resilient neuromorphic devices24,25,110. Compared to skyrmions generated in other ultrathin films, 2D skyrmions can have reduced size, high density and better tunability by external factors like electric field111 and strain112. Furthermore, by stacking 2D FM layers, the different groups of skyrmions can be formed vertically at one interface, adding a new degree of freedom to skyrmion-based spintronic devices18.

Figure 4a shows schematically an artificial synapse device based on skyrmions113. The primary component of the device is a 2D ferromagnet/HM heterostructure, which forms a nanotrack for skyrmion motion. The system replicates the operations of biological neurons and synapses through the following processes: in spike transmission mode, the pre-neuron’s spike is modulated by the synaptic device’s weight (reading through magnetoresistance), leading to the generation of a post-synaptic spike current. During the learning phase, a bidirectional charge current through the HM layer injects a vertical spin current in the FM layer, driving the skyrmions into (or out of) the post-synaptic region. This dynamic process effectively adjusts the synaptic weight, closely mirroring the potentiation and depression mechanisms observed in biological synapses. Additionally, the resolution of the synaptic weight can be adjusted based on the nanotrack width and the skyrmion size, which enriches the flexibility and tunibility of the skyrmion-based synapse. Another example for the application of 2D skyrmions is the reservoir computing24, a framework derived from recurrent neural network theory. The essence of reservoir computing is the nonlinear transformation of input into high-dimensional outputs. A physical system capable of implementing the reservoir part should possess both complex nonlinearity and memory effect (or equally hysteresis) while also exhibiting short-term properties114,115,116. Thus, magnetic skyrmions emerge as a promising candidate platform for reservoir computing. In Fig. 4b, the 2D skyrmion system performs as the “reservoir part", nonlinearly converting the one-dimensional time-series input HAC(t) to the linearly independent N-dimensional time-series outputs V(t). This design provides a guideline for developing energy-saving and high-performance skyrmion neuromorphic computing devices.

Fig. 4: Magnetic skyrmions for neuromorphic and quantum computing, with topological superconductivity searching in 2D magnetic heterostructures.
figure 4

a Schematic of the skyrmionic synaptic device. Skyrmions generated within 2D ferromagnet/heavy metal heterostructure. Bidirectional stimulus applied on heavy metal layer inducing the spin current in FM layer. b Schematic of skyrmion neuromorphic computer based on reservoir computing. Low-dimensional voltage input applied to 2D skyrmion systems yields high-dimensional magnetic output24. c Schematic of CrBr3/NbSe2 topological superconductor. dI/dV − Vbias spectrum characterized on the NbSe2 substrate (blue), the middle of CrBr3 island (red), and the edge of CrBr3 island (green)27. d Schematic of skyrmion qubit concept based on magnetic heterostructure.

Full size image

2D spintronics for quantum computing

2D magnetic heterostructures hold promise for quantum computing due to their unique electronic properties, controllability, and potential for creating novel quantum states. These interfaces can give rise to emergent quantum phenomena, broadening the horizons for quantum information processing. Notably, topological quantum computing benefits significantly from the topological protection and qubit safeguarding that 2D magnet/magnetic heterostructures can offer. An illustrative example is the observation of 2D topological superconductivity within vdW heterostructures that combine the ferromagnet CrBr3 with the SC NbSe227. As Fig. 4c shows, the monolayer CrBr3 island was grown on the NbSe2 layer. The dI/dV spectrum was characterized at different sites of this structure: NbSe2 substrate (blue), the middle of the CrBr3 island (red), and the edge of the CrBr3 (green). Among them, a peak of conductance localized at EF can be clearly seen for the edge of the magnetic island, which exhibits a hallmark of topological superconductivity for this structure. Furthermore, owing to the 2D layered structure, these edge modes can be readily accessed and manipulated using various external stimuli, including electrical, mechanical, and optical methods. This feature ushers these heterostructures for integration into topological quantum computing systems.

Another application of 2D magnetic heterostructures in quantum computing is skyrmion qubit. Unlike other proposed qubit systems such as trapped atoms, quantum dots, and photons, magnetic skyrmions offer a topologically protected structure that shows robustness against external perturbations and faults. Moreover, skyrmion qubits address challenges related to energy-efficient control and scalability, further enhancing their viability in quantum information processing. Skyrmion qubit is shown in Fig. 4d, utilizing magnetic materials platform28. In this scheme, quantum information is stored within the quantum degree of helicity or z component of magnetization, and the logical states can be dynamically adjusted through the manipulation of electric and magnetic fields.

Conclusions and perspectives

For decades, 2D magnetism has been a captivating subject, especially in the context of the Mermin-Wagner-Hohenberg (MWH) theorem, which predicts that thermal fluctuations will disrupt long-range magnetic order in 2D systems at any finite temperature, following the isotropic Heisenberg model. Extensive theoretical and experimental studies have been undertaken to unravel the causes of long-range ordering in 2D systems. Beyond magnetization, other physical properties, including polarization, which may exhibit potential long-range order within 2D systems, hold the promise of pushing the ultimate limits of Moore’s law. The inherent atomic-layer cleavability and magnetic anisotropy of 2D magnets may help mitigate spin fluctuations in the face of short-range interactions, thus paving the way for the emergence of 2D magnetism. While the duality in electromagnetism remains a profound aspect of fundamental physics, the recent discovery of 2D multiferroics further reinforces this concept, as ferroelectricity is analogous to ferromagnetism. These discoveries pose thought-provoking questions, such as the existence of an electric counterpart to the MWH theorem. Considering dipole-dipole interactions, the absence of a polarization version becomes a relevant consideration. Much like magnetic skyrmions are topological whirls in magnetization, one might contemplate the existence of an FE equivalent.

The allure of 2D magnets extends beyond fundamental physics and into the realm of novel device fabrication. Compared to sputtering, epitaxy-grown, or evaporated ultrathin films, we ought to admit the as-grown 2D materials are higher in quality, in terms of defects, disorder, or impurity levels. These 2D magnets are free from lattice mismatch problems, making high-quality interfaces possible among different 2D materials. Moreover, these 2D magnets are stacked layer by layer, ensuring an atomically smooth and dangling-bond-free surface. This is important for the uniform penetration of the electric field, leading to easy electric control of magnetic order, which is a key factor in its applications in the non-volatile industry. Significant progress has been made in comprehending the fundamental properties of 2D magnets, and the demonstration of devices is in its nascent stage. We have witnessed the expansion of the materials library and glimpsed the potential device functionalities. However, achieving integration of these materials into scalable manufacturing of functional devices requires large-area wafer-scale materials growth. Encouraging strides have been made through techniques such as MBE and atomic-layer deposition (ALD), e.g., large-scale monolayer CrBr3 with MgO as a passivation layer from ALD117 and wafer-scale FGeT on Bi2Te3 from MBE22. Challenges still exist at the device level, particularly in advancing dielectric and contact interfaces. The self-passivated nature of monolayer 2D magnets will necessitate seeding for the deposition of dielectrics through methods like ALD. It may lead to non-ideal interfaces, limiting device performance compared to the best laboratory data that employ crystalline 2D insulators like hexagonal boron nitride (h-BN). Similar challenges are encountered with electrical contacts, as they only partially conform to industry specifications and have not yet reached the level of readiness required for scalable manufacturing. Addressing such manufacturing bottlenecks will help pave the way for a significant enhancement in chip functionality, heralding a new era of 2D magnet applications characterized by increased device complexity and integration.

Moreover, in the context of neuromorphic computing, a fundamental challenge lies in improving the endurance of resistance switching. Achieving material uniformity is essential to creating massively connected device arrays capable of mimicking the hyper-connectivity and efficiency of the brain. Computational methods will be employed to guide experimental studies and optimize memristive devices for maximum performance.

For using 2D magnetic heterostructures in topological quantum computing, quantum states can benefit from topological protection, safeguarding qubits against external perturbations and errors, in contrast to standard quantum computing. Efforts in topological superconductivity and skyrmion qubits present a unique opportunity to explore the interplay between topology, magnetism, as well as electronic, magnonic, and phononic properties, potentially yielding new discoveries and insights in physics, as well as paving the way for future quantum computing devices.