Antiferromagnetic spintronics and beyond

Abstract

In this review article, we summarize some recent key results in the development of antiferromagnetic spintronics. Current-induced switching of the Néel vector orientation has now been established in a wide range of antiferromagnetic films and antiferromagnet / heavy metal bilayers, as well as current-driven motion of antiferromagnetic spin textures. The latter are particularly promising due to their small size and topological stability, but reading their magnetic state presents challenges. We also focus on materials whose compensated spin arrangements (either collinear or noncollinear) are coexistent with a spin-split band structure, enabling first-order spintronic phenomena including giant and tunneling magnetoresistance, and the anomalous Hall effect. The resulting combination of efficient electrical readout mechanisms with the advantages of a near-zero net magnetization has potential to be transformative for spintronic applications.

Introduction

Spintronics, which utilizes the quantum mechanical spin of electrons alongside their charge, introduces new paradigms for electronic devices. Ferromagnetic materials are the building blocks of spintronic technologies as their uncompensated magnetization can be easily manipulated using magnetic fields, and can be sensitively detected using giant or tunneling magnetoresistance. An antiferromagnet is characterized by a compensated spin arrangement, whose orientation is described by a Néel vector, L. The discovery of current-induced magnetic torques which act on antiferromagnetically coupled spins - the so-called Néel-order spin-orbit torques¹ - has enabled the emergence of spintronic devices with antiferromagnetic (AF) materials as their primary components, offering distinct characteristics to their ferromagnetic counterparts. Their insensitivity to external magnetic fields can be beneficial in applications where stability in the presence of external perturbations is crucial, while their intrinsic dynamics can be orders of magnitude faster than in ferromagnets. The latter can result in terahertz frequency spin rotation and switching, and domain wall speeds that are not limited by Walker breakdown^1,2,3.Though recent predictions and experimental evidence suggests domain wall speeds in ferromagnets can in some situations approach that of antiferromagnets^4,5. Furthermore, AF materials are abundant and diverse, including insulators, semiconductors, metals and superconductors.

This review summarizes some key recent results in the development of antiferromagnetic spintronics. It is not intended to be an exhaustive summary. We focus on four overlapping areas. In Section “Current-induced switching of antiferromagnetic devices” we evaluate the state-of-the-art and current understanding of electrical switching of AF domains. Section “Topological antiferromagnetic spin textures” reviews the emergence of topological spin textures in AF materials and their electrical control, which show promise as stable, nanoscale units of magnetic information. We also discuss two classes of materials which exhibit physical phenomena more commonly associated with ferromagnets, regardless of a near-perfect compensation of their magnetic moments. Section “Noncollinear antiferromagnets” discusses spintronic phenomena observed in noncollinear antiferromagnets of the form Mn₃X, with X = Sn, Ge or Pt, where the topology of the spin structure results in unique magnetic properties. For these systems, the Néel vector is not an appropriate order parameter, and higher order multipole moments must be considered⁶. Section “Altermagnetism” reviews the recently denominated field of altermagnetism, where first-order magnetic effects are observed in collinear systems which meet certain symmetry conditions.

Current-induced switching of antiferromagnetic devices

Stable orientations of local magnetic moments in both ferromagnets and antiferromagnets are separated by a magnetic anisotropy energy barrier. To produce a rotation between stable orientations, a torque must be applied to overcome this barrier. The torque can result from interaction with an either locally or globally spin-polarized current, and can have a field-like or damping-like character. The spin-polarized carriers may be injected from a ferromagnetic polarizer layer, or they may result from spin-orbit coupling. The latter presents the opportunity for magnetic memory devices which do not contain any ferromagnetic materials. For efficient current-induced switching of an antiferromagnet, the effective field driving the torque must be staggered, i.e. alternating in sign between opposite spin sublattices.

Concepts of spin-orbit torque driven switching in AF layers were first developed in around 2014¹, and were experimentally realized soon after⁷. Investigated AF devices for current-driven switching fall into two categories:

• Metallic antiferromagnets with broken parity (P) and time-reversal (T) symmetries, but combined PT symmetry, shown schematically in Fig. 1a. Key examples include CuMnAs^{7,8,9,10,11,12,13,14,15} and Mn₂Au^16,17,18,19. In these systems, due to the crystal symmetry the current induces a field-like torque of the same sign on each magnetic sublattice, i.e. a staggered effective field¹.

  Fig. 1: Schematics of current-induced switching in antiferromagnets.

  

  a Cartoon for a system where the AF sublattices are inversion partners, resulting in opposite spin currents p₁ and p₂ acting on sublattices 1 and 2 for applied current J. b Cartoon for a bilayer of a heavy metal (blue) and insulating AF (gray), where the applied current J results in interfacial spin accumulation p due to the spin Hall effect. c Typical writing current geometry, where current pulses alternately along directions indicated by red and blue arrows enable 90^∘ switching in the central section. d Geometry for reading the resulting state using transverse anisotropic magnetoresistance or spin Hall magnetoresistance, where the blue arrow indicates the probing current direction.

• Bilayers consisting of an insulating or metallic antiferromagnet plus a heavy metal, such as NiO/Pt, CoO/Pt, Fe₂O₃/Pt and MnPt/Pt (Fig. 1b)^{20,21,22,23,24,25,26,27,28,29}. Here the spin Hall effect in the heavy metal layer generates spin accumulation at the interface, resulting in a staggered damping-like torque on the AF layer.

The current pulses are typically applied in a 4-way cross geometry, with the aim of rotating the Néel vector between orthogonal directions (Fig. 1c). Electrical effects which scale with the square of the magnetization, including anisotropic magnetoresistance^7,30 and spin Hall magnetoresistance³¹, can then be used to read the resulting magnetic domain state (Fig. 1d). Simpler 2-way bar geometries have also been successfully employed^11,13,19, as well as second harmonic measurements which in principle allow 90^∘ and 180^∘ Néel vector switching to be distinguished³².

All-electrical antiferromagnetic memory devices based on these principles can show highly reproducible switching over many cycles (Fig. 2a), for current densities of 4 MA/cm², comparable to those required for switching of ferromagnetic devices⁷, and for current pulse lengths down to picoseconds¹⁰. Other notable features include a deterministic multi-level response, with potential applications for neuromorphic computing⁹, and enhanced readout ascribed to fast quenching of the AF state¹³. However, electrical readout signals in such devices may be prone to artefacts, for example due to non-magnetic inhomogeneities generated by the electrical stress from large-amplitude current pulses. Indeed, superficially similar “electrical switching" behavior is reported in non-magnetic devices subjected to current pulses on the order of 50 MA/cm^2 33,34. Moreover, transport measurements probe only average properties over the electrical contact region, and thus provide limited information on the underlying physics of antiferromagnetic domain switching phenomena.

Fig. 2: Current-induced switching in CuMnAs.

a Electrical readout after successive switching events (adapted from ref. ⁷). b Real space map of the AF domain switching with 10 μm field of view, averaged over eight pairs of pulses (adapted from ref. ⁸). c Switching of a ~1 μm AF domain by a current pulse, driven by current-induced motion of a 90^∘ domain wall (adapted from ref. ¹¹).

Synchrotron-based x-ray photoemission electron microscopy (XPEEM) provides a powerful means of directly visualizing the magnetic modifications induced by current pulses in AF devices. By tuning the incident x-ray beam to a core level absorption edge and varying its linear polarization, contrast between magnetic domains which are oriented perpendicular and parallel to the x-ray polarization direction can be resolved³⁵. Spatial resolutions below 50 nm can be achieved³⁶, and AF domain walls and structural defects can be separately resolved and correlated³⁷.

Switching of individual submicron antiferromagnetic domains in CuMnAs has been directly demonstrated using XPEEM⁸. The AF magnetic moments rotate on average into a direction perpendicular to the current pulse, consistent with the expected direction of the current-generated spin-orbit torque, but with considerable inhomogeneity due to local pinning (Fig. 2b). Current-induced motions of 90^∘ and 180^∘ domain walls in CuMnAs under modest current densities (≈4 MA/cm²) have also been directly observed using XPEEM^11,15 (Fig. 2c). In Mn₂Au, homogeneous and reversible current-induced switching has recently been demonstrated, without significant thermal activation¹⁹. For both CuMnAs and Mn₂Au, the symmetry of the domain switching observed for moderate current densities is consistent with the spin-orbit torque mechanism predicted for compounds with broken PT symmetry. With higher amplitude current pulses >10 MA/cm², the CuMnAs layer may be heated to the vicinity of its Néel temperature, resulting in a fragmentation of AF domains on lengthscales comparable to or even below the spatial resolution of the technique¹³.

Interest in this field has stimulated the development of benchtop techniques for imaging antiferromagnetic domains. For example, spin Seebeck microscopy and magneto-Seebeck microscopy have been used to investigate current-induced switching in NiO/Pt and CuMnAs, respectively^12,23. Both techniques rely on detection of thermoelectric voltages due to an optically-induced local heat gradient. Elsewhere, the large optical birefringence of NiO has been used to image AF domains in NiO/Pt structures with birefringence microscopy^26,28. Remarkably, reproducible switching of AF domains was observed even in electrically isolated regions of the NiO film, indicating that switching mechanisms beyond current-induced torques need to be considered. It was shown that the observed switching in NiO/Pt can be ascribed to a thermally induced and spatially varying strain²⁸.

The thermomagnetoelastic effect observed in NiO/Pt devices²⁸ provides an alternative mechanism to the spin-orbit torque, and the two mechanisms can act cooperatively for orthogonally applied current pulses. For CoO/Pt devices, both thermomagnetoelastic switching and spin-orbit torque driven switching were reported, depending on the current density³⁸. In CuMnAs and Mn₂Au, reversible Néel vector reorientation was observed for reversal of the current polarity, which cannot arise from thermal effects, confirming the key role of spin-orbit torque switching in these materials^11,15,19,39.

Topological antiferromagnetic spin textures

As the size of conventional magnetic logic devices, based on the electrical switching of single domains, reaches a lower limit, attention has shifted towards finding novel device architectures to continue their downsize scaling. The most promising device designs within the last two decades have focused on topologically stable, nanoscale magnetic structures as the carriers of information. These include 2-dimensional textures that resemble microscopic whirls in the magnetic order^40,41,42. The variation of spin through these textures ascribes to them a topological charge (or winding number), defined as⁴³,

$${Q}^{(k)}=\frac{1}{4\pi }\iint {{{{\bf{m}}}}}^{(k)}\cdot \left(\frac{\partial {{{{\bf{m}}}}}^{(k)}}{\partial x}\times \frac{\partial {{{{\bf{m}}}}}^{(k)}}{\partial y}\right)dxdy,$$

(1)

where m^(k) is the magnetization field and k = 1, 2 labels the two sublattices for the AF case. An AF skyrmion is composed of two topological objects with opposite winding numbers (Q^(k) = ± 1) which are strongly coupled through the AF exchange interaction⁴⁴. Q^(k) can be compactly expressed as the product between the polarity, $p={\left.-\frac{1}{2}\cos \theta (r)\right\vert }_{r = 0}^{\infty }$, and vorticity, $w={\left.\frac{1}{2\pi }\Phi (\phi )\right\vert }_{\phi = 0}^{2\pi }$, where θ and Φ are the azimuthal and polar angle components of the magnetization density, and r and ϕ are the polar coordinates^43,45. Names are given to textures according to their topological charge; notably, skyrmions and antiskyrmions, with Q^(k) = ± 1, respectively; and merons (or half-skyrmions), which have half-integer topological charge. The concept of defining magnetic textures by their topology is readily extended to 3-dimensions, with examples including more exotic Bloch points⁴⁶ and Hopfions⁴⁷. All textures with a finite topological charge are afforded topological protection. In AF materials, where dipole-dipole interactions are not a limiting factor, this enables their stability, even down to ultrasmall sizes < 10 nm⁴⁸.

Despite intensive study of topological textures nucleated and controlled in ferromagnetic materials, and a recent demonstration of their use as active components in magnetic tunnel junctions⁴⁹, their implementation into practical devices has been hindered. A considerable, unwanted effect that is exhibited in their current-driven motion is a transverse deflection, caused by a gyrotropic force originating from their topology, which reduces their suitability for racetrack device architectures. Furthermore, long-range dipole interactions inhibit skyrmion downscaling, making them susceptible to collapse at ultrasmall scales. A solution to both issues is provided by their AF counterparts, which also have benefits inherent to their compensated magnetic order: terahertz dynamics, negligible stray field, and robustness to external magnetic fields. The main alleviating factor in these AF topological textures is their compensated magnetic sublattices, which individually have a topological charge, but of opposite sign on each sublattice, so that associated gyrotropic effects cancel out^44,50.

In recent years, there has been an increase in the number of studies focused on topological textures in intrinsic AFs, but primarily in synthetic AF systems, where nucleation and stabilization mechanisms can be readily achieved using external magnetic fields and thermal effects^51,52. Despite the difficulty to nucleate and measure these textures in intrinsic systems, there have been a few breakthrough studies. In naturally occurring AF material α-Fe₂O₃, nucleation of topological textures was first demonstrated by thermal cycling using synchrotron-based imaging techniques^53,54, and nitrogen-vacancy center magnetometry⁵⁵.

In the metallic AF CuMnAs, it was shown that meron and antimeron pairs can be nucleated on a 180^∘ domain wall by an electrical pulse, as shown in Fig. 3a, b¹⁴. Subsequent pulses result in the motion of the meron-antimeron pairs along the domain wall, in the direction of the current pulse, as shown in Fig. 3di-viii. The nucleation occurs stochastically due to the combined action of joule heating and spin torque effects, while their motion can be explained by a model based on current-induced spin torques as the driving mechanism. The generated meron-antimeron pairs are homochiral as a consequence of the anisotropy and domain wall geometry, which has been shown to be crucial for their coherent current-driven motion⁵⁶. Reversible motion of these structures is achievable with current densities of (≈12 MA/cm²)¹⁴.

Fig. 3: Current-induced generation and motion of AF vortices.

A current pulse applied to the 180^∘ domain wall in a results in nucleation of a chain of vortex-antivortex pairs (b). c Numerical simulation of a vortex-antivortex pair. d Subsequent current pulses result in motion of the vortices and antivortices in the direction of the current pulse, which is indicated by the yellow arrows in the images. Figure adapted from ref. ¹⁴.

In more general AF systems, the deterministic generation of homochiral topological textures remains a focus of interest. A proposed method to achieve this is to utilize systems with a symmetry breaking interfacial Dzyaloshinskii–Moriya interaction (iDMI)⁵⁷. This has been detailed in an analytic model of α-Fe₂O₃, and has been shown to be an effective stabilization mechanism of homochiral vortices and skyrmions in synthetic AF Pt/FeCoB/Ir-based heterostructures^51,58,59.

Noncollinear antiferromagnets

A rapidly developing area in the field of antiferromagnetic spintronics is that of time-reversal symmetry breaking in compensated (or mostly compensated) spin arrangements. For example, in noncollinear antiferromagnets with a three-fold rotational sublattice transposing symmetry, the trigonal spin structure allows for breaking of the conventional t_1/2T and PT symmetries^60,61. The presence of these two symmetries imposes T-reversal symmetry in the system. When excluded however, T-symmetry may be broken. A signature of T-symmetry breaking is an anomalous Hall effect (AHE)⁶². The Hall vector, given by the off-diagonal components of the conductivity tensor h = (−σ_yz, −σ_zx, −σ_xy), transforms as a T-odd axial vector, and so is most commonly associated with the uncompensated magnetization of a ferromagnet. For a collinear antiferromagnet with T-symmetry breaking, the Hall vector depends on the orientation of the Néel vector and changes sign under its reversal (see Section “Altermagnetism”). In noncollinear systems however, the Hall vector follows the octupolar vector of the magnetic unit cell⁶³. In these systems an AHE can be measured even with complete compensation of all spins⁶⁴.

A sizeable AHE has been observed in hexagonal Mn₃Sn⁶⁵ and Mn₃Ge⁶⁶, and cubic Mn₃Pt⁶⁷, where the spins form triangular arrangements (Fig. 4). Its thermal counterpart the anomalous Nernst effect can also be readily observed in these systems^68,69, as well as other first-order magnetic effects including magneto-optical Kerr rotation^70,71 and x-ray magnetic circular dichroism⁷². The Mn₃X systems possess a small canted moment, around 2-3 orders of magnitude smaller than in typical ferromagnets, which allows a reversal of the spin configuration and resulting magnetic signals under modest magnetic fields⁶⁵. The characteristic properties of non-collinear AFs are compared to those of elemental ferromagnetic materials in Table 1.

Fig. 4: Opposite noncollinear spin arrangements showing reversal of the Hall signal.

Noncollinear spin arrangements in a the (0001) planes of hexagonal Mn₃Sn and b the (111) planes of cubic Mn₃Pt. Switching of the sublattice moments between configuration I and configuration II in these systems results in a reversal of the anomalous Hall signal and other first-order magnetic effects.

Table 1 Anomalous Hall conductivity σ_AH, anomalous Nernst conductance S_AN, Kerr rotation θ_K and uncompensated moment M in Bohr magnetons (μ_B) per magnetic atom at room temperature in the Mn₃X antiferromagnets as well as the ferromagnetic metals Fe, Co, Ni

Furthermore, the key ingredients for a practical room-temperature antiferromagnetic memory device have recently been demonstrated, namely spin polarized currents⁶¹, tunneling magnetoresistance (TMR)^73,74 and current-induced switching^75,76,77. In Mn₃Pt/MgO/Mn₃Pt, room temperature magnetoresistances of up to 100% were observed as well as exchange bias by a neighboring MnPt collinear antiferromagnetic layer⁷³. Beyond these “conventional” spintronic applications, the non-collinear antiferromagnets may offer new functionalities associated with their Weyl semimetallic band structure⁷⁸ and novel micromagnetic properties⁷⁹, providing a rich playground for physics and next-generation technology.

Altermagnetism

An altermagnet is a collinear, compensated magnetic phase in which the magnetic sublattices are connected by a rotation (proper or improper, symmorphic or non-symmorphic), but not translation or inversion symmetries.⁸⁰. The real space crystallographic rotational symmetry connecting the two opposite spin sublattices gives rise to an alternating spin-splitting of the Fermi surface in reciprocal space. Two prominent and already well studied examples are RuO₂ and α-MnTe^{81,82,83,84,85}. RuO₂ crystallises in the rutile structural, space group P42/mnm. The unconventional antiferromagnetic order (Fig. 5) corresponds to the nonrelativistic point group ²4/¹m²m¹m⁸⁰. This spin group forbids inversion and translation sublattice transposing symmetries but exhibits a transposing symmetry containing a real-space four-fold rotation ${C}_{4}{t}_{\frac{1}{2}}$. Hexagonal α-MnTe crystallises in the NiAs-type crystal structure, space group P63/mmc, where the key altermagnetic symmetry is that of a six-fold rotation, ${C}_{6}{t}_{\frac{1}{2}}$⁸⁶. It has been predicted and shown that an anomalous Hall effect can be measured in both of these systems^82,83,87. The T-symmetry breaking arises due to the anisotropy of the local magnetization densities between the two magnetic sublattices, induced by nonmagnetic atoms at locally noncentrosymmetric positions⁸¹.

Fig. 5: Key altermagnetic crystal symmetries and Fermi surfaces for RuO₂ and α-MnTe.

Crystal structures of a RuO₂ and b α-MnTe, with their real-space altermagnetic sublattice transposing symmetries. The corresponding momentum space spin-split Fermi surfaces, bounded by high symmetry directions, are depicted below.

Time reversal symmetry breaking phenomena such as the AHE, anomalous Nernst effect, Kerr rotation, X-ray magnetic circular dichroism (XMCD), spin currents, giant and tunneling magnetoresistance, and spin-torque most associated with ferromagnets have been predicted, and in some cases observed, in these compensated systems^{82,83,84,88,89,90,91}. This coexistence of ferromagnetic and antiferromagnetic like properties in a single collinear system was recently and elegantly elucidated by Šmejkal et al.^80,81. One of the recent but key experimental observations is that of their spin-split band structure via angle resolved photoemission spectroscopy (ARPES)^92,93,94,95. Krempasky et al. have confirmed a sizeable spin splitting of ~0.5 eV predicted by theory in bulk altermagnetic α-MnTe⁹³. The observed weak altermagnetic splitting predicted at k_z = 0 requires spin-orbit coupling (SOC), unlike the larger altermagnetic splitting at finite k_z that persists without. The former has been shown, with Berry phase physics, to facilitate the measured dissipationless anomalous Hall currents. While the latter, larger spin split bands, highlight the possibility of robust quantum devices in these new materials^81,82,83,91. In epitaxial thin film MnTe it was found that the band splitting persists up to 267 K, significantly lower than the bulk magnetic transition temperature⁹². There is also now evidence of spin splitting in another, high Néel temperature altermagnetic candidate CrSb⁹⁴.

A distinct feature of the AHE in these systems is its dependence on the Néel vector orientation with respect to the crystal axes. For RuO₂, the AHE is excluded by symmetry when the Néel vector is oriented along the [001] magnetic easy axis. As a consequence, the AHE shows negligible hysteresis and requires large fields for saturation (on the order of 50 T at 1.5 K)⁹⁶. In α-MnTe, the AHE is allowed when the Néel vector is oriented along a [1$\bar{1}$00] axis, which can coincide with a magnetic easy axis for certain conditions^86,97. Hence, the AHE is hysteretic and has a $\sin 3\phi$ dependence on the orientation of the Néel vector within the c-plane. As with the Mn₃X non-collinear systems, the Néel vector can be reversed using a magnetic field due to the presence of a spin-orbit-induced canted moment, which may be as small as 10⁻⁵ μ_B per Mn in the case of MnTe⁹⁸.

In RuO₂ and other materials with sufficient symmetry lowering, the anisotropy of the bands between sublattices allows for a measureable bias in the oppositely spin split isosurfaces in reciprocal space under the application of an electric field⁸¹. This allows for spin current generation as well as spin splitter like properties. An electrical spin splitter effect (SSE) with a 34^∘ propagation angle between spin-up and spin-down currents was predicted with a corresponding charge-spin conversion ratio of 28%⁹⁰. Shortly after, efficient damping-like torques and spin-to-charge conversion due to the SSE were measured in RuO₂/ferromagnet bilayers^99,100,101. Unlike the conventional spin Hall effect, the SSE does not rely on the spin-orbit interaction and is odd under time reversal. The origin of this effect is instead the anisotropic spin-split Fermi surface as shown in Fig. 5.

Outlook

Electrical writing is a key component of all magnetic memory devices. This is now well-established in antiferromagnetic materials via the Néel-order spin-orbit torque, although electrical artefacts which mimic AF domain switching have also been observed, as well as competing thermal and strain-induced switching effects. The latter provide an alternative route to realize some of the potential advantages of AF materials for spintronics.

A principal stumbling block to the technological utilization of AF materials is the lack of a practical electrical read-out mechanism. The anisotropic magnetoresistance and spin-Hall magnetoresistance are typically on the order of 0.1–1%, and effects associated with AF domain walls and spin textures may be even weaker. The need for sizeable electrical readout explains the current interest in non-collinear systems and collinear altermagnets, which combine spin-polarized transport with the compensated spin configurations. Tunneling magnetoresistance has been demonstrated using non-collinear antiferromagnets and has been predicted in all-altermagnet systems⁹¹. The spin-polarized currents and giant magnetoresistance predicted and measured in certain altermagnets also shows promise for their role in future spintronics^90,91,100. Exploiting the full potential of these materials will require improved understanding and control of their magnetic domains, as well as exploration of their dynamical properties.