Introduction

Van der Waals (vdW) magnets1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28 offer an intrinsic magnetic multilayer system with various types of magnetic ground states1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28. Among them, the interlayer antiferromagnetism with an antiparallel spin configuration across the vdW gap is known to be effectively controlled by magnetic and electric fields, pressure, or doping7,8,9,10. This leads to various spin-functionalities, particularly for antiferromagnetic (AFM) spintronics, which have gained a lot of attraction recently due to their advantageous properties over ferromagnetic (FM) spintronics, including negligible stray field, robustness against magnetic perturbation, and ultrafast spin dynamics29,30,31,32. In contrast to the so-called synthetic AFM multilayers, where FM and non-magnetic (NM) layers are stacked alternately using the thin-film deposition techniques29, vdW antiferromagnets have a highly crystalline and atomically flat interface, free from interface roughness or chemical intermixing. This may lead to highly transparent and uniform magnetic proximity interaction for better spintronic performance in the magnetic multilayers, made of vdW antiferromagnets than synthetic AFM thin films.

One key issue for vdW material-based AFM spintronics is to identify suitable candidate materials. Most of the known vdW antiferromagnets are insulating4,5,6,7,8,9,10,11,12,13,14,15,16, and their interlayer magnetic coupling is mainly through superexchange-like interaction11,12,13. Thus, the modulation of interlayer coupling usually requires structural modification9,10,11,12,13,14 and is relatively difficult as compared to synthetic antiferromagnets with the interlayer Ruderman–Kittel–Kasuya–Yosida (RKKY) coupling29. Obviously, metallic vdW antiferromagnets17,18,19,20 can be a good alternative. The conduction electrons mediate the interlayer interaction, similar to synthetic antiferromagnets, which is expected to be strongly modulated by changing the composition or the interlayer distance. Furthermore, their longitudinal or transverse conductivities are sensitive to the spin configurations33,34,35,36,37,38, offering a direct probe to the spin states even for a few nanometer-thick crystals. Despite these merits, metallic vdW antiferromagnets are rare in nature, except a few recent examples of GdTe3 and MnBi2Te4 showing a low Neel temperature TN <30 K17,18,19,20. Here, we show that an iron-based vdW material, Fe4GeTe2 with Co doping, hosts the interlayer AFM phase with TN ~226 K in a bulk and ~210 K in 8-nm-thick nanoflakes. Its spin configuration is found to be effectively controlled and read out by conduction electrons, endowing Co-doped Fe4GeTe2 with a promising role in vdW-material-based spintronics.

Results

High-T N antiferromagnetism in vdW structure

We consider iron-based metallic vdW ferromagnets FenGeTe2 (n = 3–5) as a vdW analog of the synthetic multilayer systems (Fig. 1a)21,22,23,24,25,26,27,28. The first known member, Fe3GeTe2, is experimentally identified as a ferromagnet with Tc = 220 K21,22,23,24,25, but is theoretically predicted to be interlayer AFM via RKKY-like interaction39. The interlayer AFM coupling is, however, extremely fragile to the small amount of defects or dopants39, and the AFM phase is mostly inaccessible in real compounds. An alternative candidate is Fe4GeTe2, recently identified as a high Tc metallic ferromagnet with Tc = 270 K26. Because of the relatively thick slabs containing two Fe–Fe dumbbells (Fig. 1b), it has stronger intralayer FM coupling than Fe3GeTe2, together with the interlayer FM coupling, confirmed theoretically and experimentally26. By replacing 1/3 of Fe atoms with Co atoms, however, we found that it becomes AFM, as evidenced by its temperature-dependent susceptibility χ(T) under a magnetic field H = 100 Oe along the c-axis (Fig. 1c). Hereafter (Fe,Co)4GeTe2 denotes the AFM compound (Fe1 − xCox)4GeTe2 (x = 0.33) unless the doping level x is otherwise specified. A clear cusp in χ(T), taken during the zero-field-cooling (ZFC) and field-cooling (FC), reveals AFM transition at the Neel temperature TN = 210 K. The temperature-dependent resistivity ρ(T) is metallic with an upturn at low temperatures due to Kondo scattering (Fig. 1c and Supplementary Fig. S8). The conductivity of (Fe,Co)4GeTe2 is ~3 × 105 Ω−1 m−1, which is comparable with that of the pristine Fe4GeTe226 and in the bad metal regime (Supplementary Fig. S8). These characteristics make (Fe,Co)4GeTe2 a unique vdW AFM metal with the high TN among vdW antiferromagnets (Supplementary Table S1).

Fig. 1: Crystal structure and antiferromagnetism of (Fe,Co)4GeTe2.
figure 1

a, b Schematic illustration of synthetic antiferromagnetic (AFM) multilayers and crystal structure of (Fe,Co)4GeTe2. In both cases, ferromagnetic (FM) layers with intralayer exchange coupling (Jintra) are stacked alternately and coupled by interlayer coupling (Jinter). c Temperature-dependent magnetic susceptibility χ(T) under a magnetic field H = 100 Oe along the c-axis and ρ(T) of (Fe1 − xCox)4GeTe2 for x = 0.33. A clear cusp in χ(T), taken during the zero-field-cooling (ZFC) and field-cooling (FC) reveals the AFM transition at TN ~ 210 K. d HAADF Scanning transmission electron microscopy (STEM) image of (Fe,Co)4GeTe2 (the first two ones) and Fe4GeTe2 (the last one) crystals along [1 1 0]. The STEM image with lower magnification shows that all layers are regularly arranged with a clear vdW gap (red arrow) and without stacking faults throughout whole areas. The comparison of STEM images with higher magnification of (Fe,Co)4GeTe2 (yellow box) and Fe4GeTe2 (green box) shows the almost identical atomic structure. e EELS intensity distributions of Te M, Fe L, and Co L edges within the monolayer of (Fe,Co)4GeTe2, indicating the possibilities of finding the corresponding atoms. A large intensity of Te M edge is found at the top and bottom of a monolayer as expected, whereas the intensity for Fe and Co atoms are uniformly distributed between Te layers. This infers that Co dopants successfully substitute Fe atoms as a solid solution.

Full size image

The crystal structure of (Fe,Co)4GeTe2 is the same as Fe4GeTe2 in a rhombohedral structure (space group \(R\overline{3}m\)). Scanning transmission electron microscopy (STEM) image of (Fe,Co)4GeTe2 crystal visualizes the structural units of Fe–Fe dumbbells alternately above and below the plane of Ge atoms (Fig. 1d). These Fe–Fe–Ge–Fe–Fe layers are encapsulated with Te atoms, which shows a similar structural tendency to the pristine Fe4GeTe2 (the inset in Fig. 1d). A clear vdW gap between the layers is observed without any signature of stacking change or intercalated atoms throughout a wide region. These results imply that Co atoms are dominantly substituted to the Fe sites, not in a type of the interstitial sites. In Fig. 1e, electron energy loss spectroscopy (EELS) analysis visualizes the chemical information within a monolayer, which represents that Co atoms are homogeneously doped in all the Fe sites. Co doping results in the reduction of both the in-plane (a = 4.08 ± 0.04 Å) and out-of-plane lattice parameters (c = 29.10 ± 0.28 Å), which are deducted from the selected area diffraction pattern (SADP) analysis, in good agreement with X-ray diffraction results (Supplementary Fig. S1).

Doping-dependent evolution of magnetic phases

Having established the high-TN AFM phase in (Fe,Co)4GeTe2, we focus on the systematic changes of the magnetic and electrical properties of (Fe1 − xCox)4GeTe2 single crystals with a variation of Co doping (0 ≤ x ≤ 0.39). For x = 0, the FM transition with in-plane alignment of magnetic moments occurs at Tc = 270 K, which is followed by the spin-reorientation transition to the out-of-plane alignment at TSR = 110 K (Fig. 2a)26. Co doping quickly suppresses the spin-reorientation transition, while keeping the FM transition almost intact with a nearly constant Tc for x ≤ 0.23 (Fig. 2b, c and Supplementary Figs. S2 and S3). Upon further Co doping, however, the AFM order develops from TN = 155 K, well below Tc for x = 0.26, and eventually becomes dominant with high TN up to 226 K for x = 0.39 (Fig. 2d–f). The negligible bifurcation between χ(T) curves taken during ZFC and FC is consistent with the long-range AFM phase. The saturation magnetization Msat monotonically decreases with Co doping from 7.1μB/f.u. (x = 0) to 5.5μB/f.u. (x = 0.39) due to the magnetic dilution effect (Fig. 2g). Concomitantly, the out-of-plane saturation field \({H}_{{\rm{sat}}}^{c}\) gradually increases with Co doping in the FM phases for x < 0.26. This is consistent with the changes of the magnetic anisotropic energy (K) from the easy-axis to the easy-plane types, following \({H}_{{\rm{sat}}}^{c} = 2K/M_{\mathrm{sat}}\) (Fig. 2i). Entering the AFM phase, however, \({H}_{{\rm{sat}}}^{c}\) is determined by the AFM coupling J as described by \({H}_{{\rm{sat}}}^{c}\) ≈ 2J/Msat and thus suddenly enhanced up to ~6 T for x = 0.39. In the AFM phase, the spin-flop transition is observed for H∥ab at x = 0.33, but H∥c at x = 0.39 (Fig. 2g and Supplementary Fig. S6), indicating the easy-plane and the easy-axis type spin alignments, respectively. For x = 0.39, we found that another magnetic transition occurs to the unknown phase below T = 90 K (Fig. 2f and Supplementary Fig. S2). The resulting phase diagram is summarized in Fig. 3a, which manifests that the magnetic configuration and also the magnetic anisotropy of (Fe1 − xCox)4GeTe2 are highly sensitive to Co doping.

Fig. 2: Doping-dependent magnetic phase diagram of (Fe1 − xCox)4GeTe2.
figure 2

a–f Temperature-dependent magnetic susceptibility χ(T) of (Fe1 − xCox)4GeTe2 (0.0 ≤ x ≤ 0.39), taken during zero-field cooling (ZFC) and field-cooling (FC) under H = 100 Oe with both magnetic field orientations, H∥c (solid) and H∥ab (open). g Magnetization M(H) as a function of magnetic field for (Fe1 − xCox)4GeTe2 (0 ≤ x ≤ 0.39) single crystals, for different field orientations, H∥c (solid) and H∥ab (open). All the M(H) curves were taken at T = 10 K, except those for x = 0.39 taken at T = 100 K. h Temperature-dependent in-plane resistivity ρ(T) showing a metallic behavior. i The saturation fields Hsat for H∥c (solid) and H∥ab (open) as a function of Co doping x.

Full size image
Fig. 3: First-principles calculations of (Fe1 − xCox)4GeTe2.
figure 3

a Phase diagram of spin states for (Fe1 − xCox)4GeTe2 with Curie (TC) and Neel (TN) temperatures. Four different spin states in terms of interlayer spin configurations (FM and AFM) and spin orientations (easy-axis and easy-plane) are stabilized depending on doping level and temperature. The unknown magnetic phase is found for x = 0.39 below T = 90 K. The schematic illustration of spin configurations is shown in the inset. b The saturation magnetization Msat (square symbol) and the calculated Msat (cross symbol) as a function of Co doping x. c Doping-dependent total energy difference Δ between the FM and the A-type AFM phases. The FM-to-AFM transition occurs at the critical doping level xc ~ 0.2, in good agreement with experiments. d Calculated interlayer exchange interaction Jinter with different layer distance (d) for x = 0, 0.2, and 0.5. The oscillatory dependence of Jinter(d), captured by the RKKY model, is systematically changed with Co doping. The errors in the experimental data are smaller than the size of the points.

Full size image

The first-principles calculations consistently predict the FM-to-AFM phase transition with Co doping. Total energy calculations for the FM and various AFM phases confirm that the FM phase is initially stable at low x, but eventually becomes unstable against the AFM phase at high x. The most stable AFM phase is the so-called A-type, in which all the spin moments in the whole slab of (Fe,Co)4GeTe2 are ferromagnetically aligned, but across the vdW gap, they are antiferromagnetically coupled. This is consistent with the positive Curie–Weiss temperature from the inverse susceptibility in the AFM phase, suggesting the dominant FM interaction within the layers (Supplementary Fig. S3). Furthermore, from resonant soft X-ray scattering experiments for Fe L3-edge, we observed the magnetic Bragg peak at q = (0 0 3/2) developed below TN for the crystal with x = 0.33. This confirms the A-type AFM structure, as predicted by calculations (Supplementary Fig. S7). Figure 3c shows the total energy difference ΔE = EFM − EAFM as a function of Co doping x, assuming a random distribution of Co dopants over all Fe sites. The FM-to-AFM transition occurs at the critical doping level xc ~ 0.2 (Fig. 3a). The estimated xc as well as Msat from calculations (Fig. 3b, c) is in reasonable agreement with experiments. These results indicate that the evolution of the magnetic phase with Co doping is well captured by first-principles calculations.

Detailed band structure calculations reveal that Co doping affects significantly the density-of-states (DOS) near the Fermi level. In the nonmagnetic calculations, we found a strong DOS peak in the vicinity of the Fermi level (Supplementary Fig. S9). The resulting strong Stoner instability favors the ferromagnetism within the layer, and the ferromagnetically aligned moment at each layer can be treated as a single localized Heisenberg spin, coupled through the interlayer coupling. This interlayer coupling is, however, determined by a subtle balance of pair exchange interactions between Fe/Co atoms across the vdW gap, which is sensitive to the details in the states at the Fermi level. Using total energy calculations on various interlayer spin configurations (Supplementary Fig. S10) and comparing with the classical Heisenberg Hamiltonian, we extracted the interlayer exchange interaction Jinter depending on the layer separation (d). We found the oscillatory behavior of Jinter(d), well described by the RKKY model40 (Fig. 3d). The systematic changes of Jinter(d) with Co doping, particularly between the nearest-neighboring layers, determine the stability of the AFM phase. These results contrast to the case of insulating vdW antiferromagnets, in which the changes in the interlayer coupling from the FM to AFM types require the stacking modifications9,10,11,12,13,14, and manifest the important role of conduction electrons to control the spin configurations of (Fe1 − xCox)4GeTe2.

Electrical detection of spin states

Conduction electrons are also important to probe the spin state of (Fe,Co)4GeTe2. With magnetic fields along the c-axis, normal to the preferred plane of the spin alignment, the moment at each layer of (Fe,Co)4GeTe2 gradually rotates until its mean direction is parallel to the field at \({H}_{{\rm{sat}}}^{c}\). This out-of-plane spin canting can be monitored by the anomalous Hall effect (AHE) due to sizable spin–orbit coupling in (Fe,Co)4GeTe2. As shown in Fig. 4a, the Hall resistivity ρyx is dominated by the anomalous contribution \({\rho }_{yx}^{\mathrm{A}}\), i.e., ρyx ≈ \({\rho }_{yx}^{\mathrm{A}}\), and thus the transverse conductivity σyx = ρyx/(\({\rho }_{xx}^{2}\) + \({\rho }_{yx}^{2}\)) with a variation of magnetic field is nicely scaled with M(H). Moreover, the scaling factor SH = σyx/M is found to be almost independent of temperature (Supplementary Fig. S5), and thus σyx(H, T) can quantitatively measure the out-of-plane component of net magnetization M(H, T). For example, the magnetic susceptibility, estimated from χc(T) ∝ σyx(H)/H, allows for experimentally determining TN in nanoflakes (Fig. 5).

Fig. 4: Electrical detection of the spin states.
figure 4

a The transverse conductivity σyx(H) as a function of magnetic fields along the c-axis, taken at various temperatures. The σyx(H) data are nicely reproduced by the field-dependent magnetization M(H) (black solid line) with a scaling factor SH ≈ 0.3 V−1, following the linear relation of σyx(H) = SHM(H). b Magnetoresistance Δρ(H)/ρ(0) under in-plane magnetic fields H, parallel (open) or perpendicular (solid) to the current I along the a-axis. The low field spin–flop transition field \({H}_{{\rm{sf}}}^{ab}\) and the high field saturation field \({H}_{{\rm{sat}}}^{ab}\) are indicated by the arrows. c Spin configurations with different relative orientations of the magnetic field H and the current I. At low H, the antiferromagnetically coupled spins are aligned perpendicular to H, either H∥I (a) or H⊥I (b). At high H, the saturated spins are aligned parallel to H, either H∥I (c) or H⊥I (d). d Anisotropic magnetoresistance ΔρAMR as a function of temperature and in-plane magnetic field. The low-field and high-field AMR, determined by the relative orientation of Neel vector and the saturated magnetization against the current direction, respectively, which results in a sign change.

Full size image
Fig. 5: Thickness-dependent magnetic phase diagram.
figure 5

a–d Magnetic field-dependent transverse conductivity σyx(H) at various temperatures of (Fe1 − xCox)4GeTe2 nanoflakes (x = 0.33) for H∥c in 16L (a), 7L nanoflakes (b), and (Fe1 − xCox)4GeTe2 nanoflakes (x = 0.39) for H∥c in 20L (c) and 9L nanoflakes (d). The σyx(H) curves at low temperatures show typical coercivity in FM phases (b–d). e Temperature dependence of the magnetic susceptibility χc(T) for (Fe1 − xCox)4GeTe2 crystal (x = 0.33), estimated from the low-field slope of σyx(H) at each temperature (open circle) or from the difference between σyx(T) curves taken at H = ±0.1 T for H∥c (solid circle). Magnetic transition temperature TN and TC are indicated by the arrows, together with the characteristic temperature T*, determined by a broad hump in χc(T). f Thickness-dependent magnetic phase diagram of (Fe1 − xCox)4GeTe2 crystal (x = 0.33) with characteristic temperatures of TN, T*, and Tc. g Temperature dependence of the magnetic susceptibility χc(T) for (Fe1 − xCox)4GeTe2 crystal (x = 0.39) (open circle), bulk magnetization with H = 0.1 T for H∥c (solid symbol). Characteristic temperature T0 is indicated by the arrows at the inflection of χc(T) at low temperature. h Thickness-dependent magnetic phase diagram of (Fe1 − xCox)4GeTe2 crystal (x = 0.39) with characteristic temperatures of TN, T0, and Tc. The errors in the experimental data are smaller than the size of the points.

Full size image

The orientation of the staggered magnetization in the plane, i.e., the Neel vector in the plane can also be effectively tracked by the electrical conductivity. Figure 4b shows the field-dependent magnetoresistance Δρ(H)/ρ(0) of (Fe,Co)4GeTe2 crystal under different field orientations in the plane, H∥I and H⊥I against a current I∥a at various temperatures. Two characteristic fields, \({H}_{{\rm{SF}}}^{ab}\) and \({H}_{{\rm{sat}}}^{ab}\), are identified from clear kinks in both Δρ(H)/ρ(0) curves. The low field \({H}_{{\rm{SF}}}^{ab}\) corresponds to the spin–flop transition, at which the Neel vectors of all domains are fully aligned perpendicular to the in-plane field33,34,35,36,37,38. The Neel vector L is rotated by 90∘, from L⊥I to L∥I, by switching the in-plane magnetic field from H∥I to H⊥I (Fig. 4c). This leads to a difference in magnetoresistance due to spin–orbit coupling, which is known as the anisotropic magnetoresistance (AMR) \({{\Delta }}{\rho }_{{\rm{AMR}}}^{L}\) = ρL∥I − ρL⊥I. In (Fe,Co)4GeTe2, \({{\Delta }}{\rho }_{{\rm{AMR}}}^{L}\) reaches ~0.3%, comparable with other AFM metals33,34,35,36. With further increasing in-plane field, the moments are canted towards the field, until they are fully aligned at the saturation field \({H}_{{\rm{sat}}}^{ab}\). In this case, the AMR between the cases of M∥I or M⊥I, \({{\Delta }}{\rho }_{{\rm{AMR}}}^{M}\) = ρM∥I − ρM⊥I is also expected as found in ferromagnets41, 42. In (Fe,Co)4GeTe2, \({{\Delta }}{\rho }_{{\rm{AMR}}}^{L}\) and \({{\Delta }}{\rho }_{{\rm{AMR}}}^{M}\) are comparable in size and opposite in sign, inducing the sign cross of ΔρAMR as summarized in Fig. 4d. Therefore, the AMR allows the electrical access to the orientation of the Neel vector or the saturated magnetization in the plane. Therefore, by combining AHE and AMR we can effectively read out the spin state of (Fe,Co)4GeTe2.

Thickness-dependent magnetic phases

Finally, we discuss the thickness control of the magnetic state of (Fe,Co)4GeTe2. Owing to the weak vdW interlayer coupling, we obtained nanoflakes with thickness d down to seven layers (7L) for x = 0.33 and nine layers (9L) for x = 0.39, using mechanical exfoliation. In both cases, the in-plane resistivity ρ(T) shows the metallic behavior (~10−3 Ωcm) with a low-temperature upturn due to Kondo scattering, similar to the bulk case (Supplementary Figs. S12 and S13). In nanoflakes, we used the transverse conductivity σyx(H, T) to monitor the field and temperature-dependent magnetization M(H, T), as done in the bulk case (Fig. 4a). The resulting susceptibility χc(T) of nanoflakes (Fig. 5e, f and Supplementary Figs. S12 and S13) exhibits a clear kink, indicating the AFM transition at TN in both cases with x = 0.33 and 0.39.

The detailed thickness-dependent phase diagram is different with Co doping level x. For nanoflakes with x = 0.33, T  decreases gradually from the bulk value with lowering thickness. In addition, in nanoflakes, we found a broad hump developed around T* in χc(T), which shifts to low temperatures with reducing thickness (Fig. 5e). This nonmonotonous temperature dependence of χc(T) implies that spin moments are not fully frozen by dominant AFM interaction in nanoflakes, due to additional competing FM interaction. With further reducing thickness to 7L, this competing FM interaction stabilizes the long-range FM order, evidenced by clear magnetic hysteresis of σyx(H) with a coercive field Hc = 0.21 T at T = 2 K (Fig. 5b). The Hc gradually decreases with increasing temperature up to Tc ≈ 25 K (Supplementary Fig. S12). The corresponding χc(T) exhibits no signature of AFM ordering but only a broad peak at T* ~ 40 K. The resulting phase diagram (Fig. 5g) reveals that thickness tuning offers another effective means to tune the interlayer FM and AFM interactions, even though the 2D limit is yet to be reached. It has been well established that vdW crystals expand along the c-axis by ~0.3–0.7%, 43, 44 when they are thinned to be tens of nanometer thick. Considering that the c-axis lattice constant shrinks by ~0.6% with Co doping of x = 0.39, this thinning-induced swelling affects significantly the magnetic ground state near the FM–AFM phase boundary.

In nanoflakes with higher Co doping of x = 0.39, located deep inside the AFM phase of the phase diagram (Fig. 3a), the AFM phase is expected to be more stable than the nanoflakes with x = 0.33. As shown in Fig. 5, this is indeed the case. Upon reducing the thickness down to 9L, we observed similar σyx(H) curves with those of the bulk (Fig. 2g and Supplementary Fig. S5d). TN is also reduced slightly to ~210 K from its bulk value of 226 K, and the low-temperature transition at T0 is also maintained without significant change. At low temperatures, we observed the magnetic hysteresis (Fig. 5c, d and Supplementary Fig. S13), indicating the FM phase, as found in the case of x = 0.33. The resulting thickness-dependent phase diagram for x = 0.39 reveals that the AFM phase in the intermediate temperature range between T0 and TN is stable in nanoflakes. The spin–flop transition field HSF, however, is systematically reduced with lowering thickness (Supplementary Fig. S13f). These results clearly demonstrate that the stability of the AFM phase and the magnetic anisotropy can be controlled by thickness and doping levels in (Fe1 − xCox)4GeTe2.

Discussion

Our observations unequivocally show that (Fe,Co)4GeTe2 is an intrinsic high-TN AFM multilayers. Although its TN is still below room temperature, an order of magnitude enhancement of TN, as compared to other metallic vdW antiferromagnets17,18,19,20, is achieved by tuning the vdW interlayer coupling between the strongly FM layers of Fe4GeTe2. This approach can be applied to other recently discovered high-Tc vdW ferromagnets26,27,28, 42 to possibly realize the room temperature antiferromagnetism. Furthermore, (Fe1 − xCox)4GeTe2 hosts at least four different spin states in terms of interlayer spin configurations (FM and AFM) and spin orientations (in-plane and out-of-plane) while keeping the same stacking structure (Fig. 3a). The switching between these states, demonstrated using doping, magnetic field, and thickness control, can be more effectively achieved in vdW heterostructures through, e.g., the exchange-spring effect45, 46 with an adjacent FM vdW layer or the current-induced spin–orbit torque34, 37 with large SOC materials, as done in metal spintronics. The intimate coupling of the spin states to electrical conduction in (Fe, Co)4GeTe2 (Fig. 4a, b) ensures the electrical readout of the spin state. These strong controllability and readability on the spin states are the manifestation of itinerant magnetism, highly distinct from insulating vdW magnets. We, therefore, envision that metallic vdW antiferromagnets, including (Fe,Co)4GeTe2 in this work, will enrich the material candidates and the spin functionalities for all vdW material-based spintronics.

Methods

Single-crystal growth and characterization

Single crystals were grown by a chemical vapor transport method using pre-synthesized polycrystalline sample and iodine as a transport agent. The obtained single crystals had plate-like shape with a typical size of ~1 × 1 × 0.04 mm3. The high crystallinity of single crystals was confirmed by X-ray diffraction (Supplementary Fig. S1). From the energy-dispersive spectroscopy measurements, we confirmed a systematic variation of Co doping x, in the presence of Te deficiency by ~10% and excess of the total (Fe, Co) content by ~5%, similar to pristine Fe4GeTe226. Magnetization was measured under magnetic field along the c-axis or ab-plane using a superconducting quantum interference device magnetometer (MPMS, Quantum Design) and vibrating sample magnetometer option of Physical Property Measurement System (PPMS-14T, Quantum Design). The in-plane resistivity and the Hall resistivity were measured in the standard six probe configuration using a Physical Property Measurement System (PPMS-9T, Quantum Design).

Device fabrication

Using mechanical exfoliation in the inert argon atmosphere (H2O < 0.1 p.p.m., O2 < 0.1 p.p.m.), we obtained nanoflakes of (Fe,Co)4GeTe2 on top of Si/SiO2 substrate, pre-treated by oxygen plasma (O2 = 10 s.c.c.m., P ~ 100 mTorr) for 5 min to remove adsorbates on the surface. The exfoliated crystal is then subsequently covered by a thin h-BN flake to prevent oxidation during device fabrication. Typically (Fe,Co)4GeTe2 flakes are of several tens of μm2 in the area and down to ~7 nm (~7L) in thickness, estimated by the atomic force microscopy measurements (Supplementary Fig. S11). To make electrodes for transport measurements, we employed an electron beam lithography technique, using poly(methyl methacrylate)-positive resist layer, which was spin-coated and dried in vacuum at room temperature. After etching, the patterned area of the covered h-BN flake with CF4 plasma, Cr(10 nm)/Au(50 nm), is deposited on the exposed surface of the nanoflake.

STEM and EELS

STEM samples were made along [1 1 0] projections, which show the most distinguishable atomic structures. Samples were prepared with dual-beam focused ion beam systems (Helios and Helios G3, FEI). Different acceleration voltage conditions from 30 to 1 keV were used to make a thin sample with less damages. The subsequent Ar ion milling process was performed with low energy (PIPS II, Gatan Inc., USA). The atomic structure was observed using a STEM (JEM-ARM200F, JEOL, Japan) at 200 kV equipped with a fifth-order probe corrector (ASCOR, CEOS GmbH, Germany) at Materials Imaging & Analysis Center of POSTECH in Republic of Korea. The electron probe size was ~0.8 Å, and the high-angle annular dark-field detector angle was fixed from 68 to 280 mrad. The SADP was obtained using a TEM (JEM-2100F, JEOL, Japan) at 200 kV equipped with a spherical aberration corrector (CEOS GmbH, Germany). The raw STEM images were compensated with ten slices by SmartAlign and processed using a band-pass difference filter with a local window to reduce background noise (SmartAlign and Filters Pro, HREM Research Inc., Japan). For EELS analysis, we utilized another STEM (JEM-2100F, JEOL) with a spherical aberration corrector (CEOS GmbH) equipped with an EEL spectrometer (GIF Quantum ER, Gatan, USA). The used probe size was ~1.0 Å under 200 kV and the chemical analysis was performed by using a Spectrum Image via the STEM mode. The obtained spectral data from a spectral image were filtered to intensify the Fe-, Co-, and Te-edge signals by MSA (Multivariate Statistical Analysis, HREM Research Inc., Japan).

Scanning tunneling microscopy

The surface of (Fe,Co)4GeTe2 was characterized by the scanning tunneling microscopy (STM) on a single crystal. The single crystal is first cleaved in a high vacuum (~1 × 10−8 torr) to obtain the clean surface and then is transferred to ultra-high vacuum (~1 × 10−11 torr) for the STM measurements at 77 K.

First-principles calculations

Electronic structure calculations were performed using a full-potential linearized augmented plane wave method, implemented in WIEN2K package 47. Experimental lattice constants of (Fe,Co)4GeTe2 are used and exchange-correlation potential is chosen to the generalized gradient approximation of Perdew–Berke–Ernzerhof48. Assuming that Co is doped at all Fe sites homogeneously throughout the Fe4GeTe2 layer, the virtual crystal approximation is considered in the DFT calculation of (Fe,Co)4GeTe2.

Soft X-ray scattering

Resonant soft X-ray scattering experiment was carried out in 6A MPK MeXiM beamline, PAL. The single crystals were cleaved in the air and thereafter immediately transferred to the chamber with the ultra-high vacuum pressure of ~8 × 10−9 torr. Photon energy was selected near the Fe L3-edge absorption energy value at 707.5 eV, because q = (0 0 3/2) peak has a different energy profile with (0 0 3) Bragg peak.