Progress and Prospects in Metallic Fe_xGeTe₂ (3 ≤ x ≤ 7) Ferromagnets

Abstract

Thermal fluctuations in two-dimensional (2D) isotropy systems at non-zero finite temperatures can destroy the long-range (LR) magnetic order due to the mechanisms addressed in the Mermin-Wanger theory. However, the magnetic anisotropy related to spin–orbit coupling (SOC) may stabilize magnetic order in 2D systems. Very recently, 2D Fe_xGeTe₂ (3 ≤ x ≤ 7) with a high Curie temperature (T_C) has not only undergone significant developments in terms of synthetic methods and the control of ferromagnetism (FM), but is also being actively explored for applications in various devices. In this review, we introduce six experimental methods, ten ferromagnetic modulation strategies, and four spintronic devices for 2D Fe_xGeTe₂ materials. In summary, we outline the challenges and potential research directions in this field.

1. Introduction

The Mermin-Wanger theory [1,2] asserts that thermal fluctuations occur in 2D isotropy systems at non-zero finite temperatures, which destroy the long-range magnetic order (LRMO). Specifically, exchange interactions alone should not generate magnetic order in 2D systems, and magnetic anisotropy [3,4,5] is also needed to maintain the LRMO. Surprisingly, it was found experimentally that low-temperature long-range ferromagnetic order (LRFO) can exist in the Cr₂Ge₂Te₆ monolayer [4] and CrI₃ monolayer [5,6]. Soon after, a vast range of 2D magnetic systems, including metallic (Fe₃GeTe₂ (FGT) [7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23]), semiconductors (Cr₂Ge₂Te₆ [4,24,25,26,27,28,29,30,31,32], CrI₃ [5,33]), and topological insulators (MnBi₂Te₄ [34]), were successively implemented to promote the development of spintronics.

Recently, Fe_xGeTe₂ (3 ≤ x ≤ 7) has received intense attention as a metallic and high Curie temperature (T_C) ferromagnet. Six synthesis methods, including solid-state reaction (SSR) [35,36], chemical vapor transport (CVT) [8,13,37], the flux method [11,21,38,39,40,41,42,43,44,45], exfoliation [14,15,34,46,47,48,49,50], chemical vapor deposition (CVD) [51,52], and molecular beam epitaxy (MBE) [7,53,54,55,56,57,58], have been used to attempt to obtain wafer-scale Fe_xGeTe₂ (3 ≤ x ≤ 7) materials with room-temperature ferromagnetism (RTFM). However, the T_C of the MBE-prepared Fe_xGeTe₂ (3 ≤ x ≤ 7) samples (see Figure 1) ranges from 390 to 530 K, with FGT (T_C ≈ 400 K) [56], Fe₄GeTe₂ (F4GT; T_C ≈ 530 K) [59], and Fe₅GeTe₂ (F5GT; T_C ≈ 390 K) [60]. Furthermore, RTFM has also been tuned with ten strategies: Fe stoichiometry [9,39,51,59,61,62,63,64,65], strain engineering [46,48,66,67,68,69,70,71,72,73,74,75], hydrostatic pressure [76,77,78,79,80,81], light control [53,82], electrical control [83,84], proximity effects [56,57,85,86,87,88,89], doping engineering [14,20,38,43,44,62,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106], intercalation [107,108] or irradiation [109], twisting [110,111], and patterning [16]. So far, twisting (see Figure 1) has only regulated the magnetic order theoretically and has not been achieved experimentally. Moreover, four typical devices have also been fabricated based on FGT, magnetic tunnel junctions (MTJ) [112,113,114,115], tunneling spin valves [18,99,116,117], nonlocal spin valves [118], and spin–orbit torque devices [20,119], in order to enrich their physical properties and develop their spintronic applications.

In this review, we introduce the developments in and structures of metallic Fe_xGeTe₂ ferromagnets. Subsequently, we summarize six experimental methods, as shown in Figure 1; the early samples prepared using SSR, CVT, and the flux method are mainly bulk single crystals. In addition, magnetic skyrmions in Fe_xGeTe₂ are summarized. Finally, we outline the challenges and potential research directions in this field.

2. Crystal Structure of Ferromagnetic Fe_xGeTe₂

The Fe₃GeTe₂ monolayer [46] comprises five atomic layers, as shown in Figure 2A. Specifically, Te atoms are located in the bottom and top layers, while Fe (I) atoms are located in the second and fourth layers. Notably, the intermediate layer is composed of Fe (II) atoms and Ge atoms. The local magnetic moments of Fe atoms for the FGT monolayer were determined by DFT-LDA (density functional theory–local density approximation) to be 1.723 µ_B and 1.005 µ_B, with the out-of-plane direction being its easy axis. They may be related to the several partially occupied d-bands passing through the Fermi level. In addition, the number ratio of Fe³⁺ to Fe²⁺ in 2D Fe_xGeTe₂ [64] is related to the x value, as shown in Figure 2B. When the value of x is 3, the ratio of Fe³⁺/Fe²⁺ is 2:1. However, when the x value is 5, only Fe³⁺ is present (Figure 2C).

3. Synthesis of Metallic Fe_xGeTe₂ with FM

3.1. Solid-State Reaction (SSR)

The solid-state reaction (SSR) is an experimental method for preparing bulk FGT crystals. As early as 2006, Deiseroth et al. [35] successfully prepared FGT crystals with hexagonal plates using SSR, which exhibited novel air stability and black metallic properties. Through magnetic testing, it was found that below 230 K, the crystals exhibited FM. Meanwhile, above 230 K, they exhibited Curie–Weiss paramagnetic behavior. After increased annealing, black Fe_3−δGeTe₂ (0 < δ < 0.3) polycrystalline powders could be easily obtained with SSR. Fe1 and Ge1 atoms are shown in detail in different coordination environments in Figure 3A,B, and their unit cell consist of two layers. The lattice parameters increase monotonically with decreasing δ (it represents the degree of iron deficiency in FGT) (Figure 3C), but, when δ exceeds 0.3, FeTe₂ will appear as an impurity phase. Its magnetic phase transition temperature (Figure 3D) is about 240 K. Furthermore, its saturation behavior (see inset in Figure 3D) slows down in high magnetic fields, which is different from ordinary ferromagnets.

In order to obtain large quantities of high-quality FGT single crystals, Li et al. [120] designed a new experimental method of solid-phase sintering followed by recrystallization (Figure 3E–G). The as-grown plate-like sample (~10 g) is a layered single crystal with a smooth and complete surface, and its size can reach up to 8.5 mm. By intercalating sodium into as-grown FGT, Weber et al. [107] raised its T_C to 350 K. After intercalation, the sample retained obvious layered features, with edge lengths of a grain size ranging from 10–50 μm.

3.2. Chemical Vapor Transport (CVT)

One main difference from SSR is that CVT often uses iodine [8,13,14,15,17,37,39,61,119] or TeCl₄ [12] as the transport agent, as shown in Figure 4. However, the samples obtained via SSR and CVT were both bulk single crystals, as shown in the inset of Figure 4A,B. Previous studies have mainly focused on the magnetic microstructures of quasi-2D FGT. Based on the prediction that FGT monolayer could be mechanically exfoliated [46], soon after, Chu et al. [15] and Zhang et al. [14], respectively, obtained FGT monolayer samples with the assistance of Au film and Al₂O₃, respectively. Actually, Zhang et al. [121] exfoliated FGT monolayer from the most possible cleaving planes (001), with a thickness of 1.75 nm and a nearest neighbor atomic spacing of 0.338 nm, which was highly consistent with the lattice constant (a = 0.399 nm; c = 1.63 nm) of the FGT crystal. However, thin layer FGT was highly prone to deteriorate in air, and the device fabrication processes (Figure 4C–E) needed to be carried out in a glove box [49]. Notably, many novel physics-related effects, such as patterning-induced RFTM [16], gate-tunable FM [14], and layer-dependent FM [15], have been discovered.

3.3. Flux Growth

The flux method [122,123,124] is commonly used to prepare single crystals. For example, Canfield et al. [122,123] grew a wide variety of single-crystal binary or ternary intermetallic compounds from molten flux solutions. However, the thickness and lateral size of the samples could not be accurately controlled, with mechanical exfoliation still required to obtain thinner samples when fabricating FGT devices.

Recently, Gong et al. [44,45] proposed a universal flux-assisted growth (FAG) method (Figure 5A,B) to synthesize Fe_xGeTe₂ and M_yFe_5−yGeTe₂ (M = Co, Ni) nanosheets on various substrates (Figure 5C). In addition, the sample thickness and lateral size (Figure 5D) of FGT could be precisely controlled by the growth temperature (Figure 5E) or cosolvents (Figure 5F). Although the FGT samples with a thickness of 5–10 nm (Figure 5G) were prepared on various substrates, in order to obtain atomically thin materials (ATMs), a confinement environment must be provided through two substrates. Up to 80 layered and non-layered ATMs [45] have also been successfully synthesized using FAG, which provides a new strategy for preparing wafer-scale 2D materials.

3.4. Exfoliation

3.4.1. Mechanical Exfoliation

Conventional mechanical exfoliation [125,126] can cleave thin FGT flakes onto SiO₂/Si substrates, but its thinnest thickness is around 4.8 nm. After depositing Au onto SiO₂/Si substrate, thinner samples can be obtained, and the Au substrate improves the yield to grow various thin layers of materials, including graphene [127,128], MoS₂ [47,129,130,131,132], WSe₂ [47,129,133], Bi₂Te₃ [129], and FGT [15,47]. Nevertheless, only a small amount of material can be exfoliated to a monolayer, which hinders the development of 2D magnetic materials. Notably, an Al₂O₃-assisted exfoliation method was also designed to produce monolayer FGT [14] and MnBi₂Te₄ [34] single crystals. When the sample was thinned from bulk to a monolayer, its T_C decreased from 180 K to 20 K.

3.4.2. Liquid-Phase Exfoliation

Although many methods including SSR [107], CVT [14,15,17,19,61,119,121,134], flux [39,43,44,45], and MBE [7,53,54,55,56,57,58,59,60] have been used to prepare 2D FGT, an economical method for the large-scale preparation of few- or single-layer FGT nanoflakes is still lacking. As a typical example, Ma et al. [50] developed three-stage sonication-assisted liquid-phase exfoliation (TS-LPE) (Figure 6A) to produce large semiconductive FGT nanoflakes. After ball milling, the sample size and thickness (Figure 6B,C) are reduced by the milling time (Figure 6G), exposing more boundaries. Stirring causes the interlayer spacing to expand (Figure 6C,D), weakening the interlayer force to facilitate detachment and obtain high-integrity nanoflakes. In addition, XRD analysis [135] (Figure 6H,I) reflects the evolution of interlayer spacing. The expansion of interlayer spacing causes the FGT unit cell to move away from the equilibrium state in the c-direction (Figure 6J), making them unstable and prone to spall. In practice, the oxidation on the surface layer altered the electronic structure of the FGT system, making the FGT sample semiconductive and different from the metallic FGT prepared using other methods.

3.5. Chemical Vapor Deposition (CVD)

So far, researchers have mainly used CVT to prepare 2D magnetic bulk single crystals, which are then exfoliated into atomic layers to prepare devices. However, poor control of the number of layers and a limited sample size have hindered the development of 2D magnets. As a typical example, Liu et al. [51] designed a confined space chemical vapor deposition (CS-CVD) method for preparing 2D FGT or F5GT ferromagnets. They found that the optimal growth temperature was 570–580 °C, with an optimal distance of 10 cm between the Fe/Ge precursor and the Te precursor. When the thickness of the F5GT flakes changed from 4 nm to 1 nm, the T_C value decreased by 100K. Very recently, Liu et al. [52] also introduced a general competitive-chemical-reaction-controlled CVD method for producing FGT crystals. The sample was a single layer with a grain size of ~50 μm.

3.6. Molecular Beam Epitaxy (MBE)

Wafer-scale single crystalline FGT thin films were grown on various substrates using the molecular beam epitaxial (MBE) [7,54,55,56,57,136] technique. After heterointegration with the topological insulator Bi₂Te₃ (Figure 7A), the T_C of FGT can be increased to 400 K. This enhancement may be related to the interface exchange coupling. Remarkably, when the thickness of F4GT decreases, its T_C is increased from 270 K to 530 K (Figure 7B). For F4GT thin films with a thickness of 10 nm, increasing the dosage of Fe can enhance their T_C (Figure 7C). Although MBE can alone prepare wafer-scale Fe_xGeTe₂ (3 ≤ x ≤ 7) materials with RTFM, this method requires a high vacuum environment, which makes it expensive and limits its industrial applications.

4. Controlling FM in Metallic Fe_xGeTe₂

4.1. Fe Stoichiometry

In this research area, the earliest discovery was that the FM in polycrystalline FGT bulk structures [9] was related to the Fe content (Figure 8). The higher the Fe content, the larger the lattice constant of the a-axis and the smaller the lattice constant of the c-axis (Figure 8A). Single crystal samples show similar results to the polycrystalline samples. Moreover, the T_C (Figure 8B) and M_S decreased with the decrease in Fe content. Subsequently, ferromagnetic F4GT [44,59,61] and F5GT [39,42,44,51,62,137,138,139,140,141,142,143,144] materials were also obtained in experiments.

However, most previous reports have focused on FGT materials with a single Fe stoichiometry, and there have been few studies on Fe_xGeTe₂ materials using the same experimental method. In addition, theoretical calculations [63] revealed that as the Fe content increased, the interlayer gap gradually increased, and the magnetic anisotropy of its monolayer changed from out-of-plane (FGT) to in-plane (F4GT and F5GT).

Although previous reports have made significant progress in 2D Fe_xGeTe₂ systems, the mediation of magnetic anisotropy and their magnetic nature remains unresolved. Very recently, Liu et al. [64] preliminary reported a valence-dependent magnetic exchange model to explain the complex magnetic phase in Fe_xGeTe₂ systems. Furthermore, the magnetic moment and MAE (Figure 9A) were almost linearly correlated with Fe²⁺/Fe³⁺. Specifically, Fe³⁺ had a greater impact on magnetism, reducing the magnetic anisotropy energy in F5GT. Based on MAE and J (it represents the exchange coupling constant), the T_C could be estimated using the 2D Heisenberg model. When x was greater than 4, the T_C was much higher than RT (Figure 9B).

Moreover, the results obtained via different calculation methods [65] exhibit significant differences, especially when compared with experimental results. As x increases, its easy axis direction (shown by the black arrows) and the highest exchange interaction change from out-plane to in-plane (Figure 9C–F). Moreover, there were significant differences in the magnitude of the exchange interaction obtained via different calculation methods (GGA + DMFT, GGA, and GGA + U); specifically, the results calculated using GGA + U were overestimated. In addition, MAE exhibits a similar evolution from out-plane to in-plane. However, for different F5GT (UUU or UDU) configurations, the calculation results varied significantly. It was obvious that the T_C calculated via GGA + DMFT was underestimated, while the result calculated by GGA was overestimated, compared to Figure 9B.

4.2. Strain Engineering

Strain engineering is an efficient strategy for modulating the FM of 2D materials [67,68,145]. However, previous theoretical works have focused on applying strain to FGT supercells by changing the lattice constants [46,70,71,73,146] and calculating the exchange coupling, magnetic anisotropy, and magnetic moment of strain through ab initio DFT. Furthermore, the T_C could be estimated according to mean field theory (MFT) [10,61,65,147,148,149], random phase approximation (PRA) [147,149], or Monte Carlo (MC) [148,149,150,151] simulation. Recently, Miao et al. [48] and Yan et al. [72] loaded FGT nanoflakes into a three-point-bending experimental setup and applied uniaxial tensile strain to the sample on a polyimide (PI) or polyvinyl alcohol (PVA)/polyethylene terephthalate (PET) flexible polymer substrate by moving the needle. Moreover, the magnitude of the applied strain could be calculated using the following formula [48,72,152]:

$$\epsilon =\frac{T}{2R}$$

(1)

where T and R are the film thickness and bending radius, respectively. Surprisingly, after 0.32% strain was applied, the coercivity increased by 150% [48], a far greater increase than the improvement in H_C generated by other traditional magnetic materials [153]. Most notably, its T_C could be increased to 400 K [72] via uniaxial strain, further promoting the development of wearable spintronic devices with low-energy consumption.

4.3. Hydrostatic Pressure

Tuning the exchange coupling and magneto-crystalline anisotropy by applying hydrostatic pressure is another commonly used method for regulating 2D magnetism, which has been achieved in Cr₂Gr₂Te₆ [154,155], CrI₃ [33,156], and FGT [77,78,79] systems (Figure 10). The ferromagnetic evolution of FGT nanosheets under different pressures can be revealed through in situ magnetic circular dichroism (MCD) spectroscopy (Figure 10B). Furthermore, the magnetic hysteresis loop at 30 K exhibited a rectangular shape below 7 GPa, while its loop presented an eight-shaped skewed shape above 7.3 GPa. Moreover, T_C increases as the pressure further decreases (Figure 10C), which may be related to the strengthening of the exchange interactions.

As another typical example, T_C and the magnetic moment also increased with the decrease in the hydrostatic pressure (Figure 10D,E). It is evident that the increasing pressure reduced the length of the Fe–Fe bond, which inhibited magnetization through modification of the exchange interactions. In addition, a monotonic relationship between T_C, the magnetic moment, and pressure was also found in the Fe-deficient FGT sample [77], similar to the FGT system. Overall, pressure modifies the metallic form of Fe_3−xGeTe₂ to its nonmetallic form.

4.4. Light Control

The continuous modulation of monolayer transition-metal dichalcogenides (TMDs) without intrinsic magnetism, including MoS₂ [157], WS₂ [157], and WSe₂ [158], has been achieved using the optical approach. Recently, Tengdin et al. [82] demonstrated that spin polarization was transferred from Mn sublattices to Co on the Heusler compound Co₂MnGe via femtosecond laser pulse, which is closely related to the wave function of electrons before and after being excited by light. The ultrafast spin transfer caused by the instantaneous incident light on the material does not only occur in Co₂MnGe, but is also a common feature of many materials. Notably, Xu et al. [53] reported that the magnetic anisotropy energy (MAE) and T_C were mediated with a femtosecond laser pulse, as shown in Figure 11. The optical doping effect alters the electronic structure of FGT, thereby affecting exchange interactions, T_C, and MAE. According to Figure 11B, the T_C of FGT was estimated to be ~200 K. Under the excitation of a femtosecond laser, electrons transitioned from an occupied state to an unoccupied state, causing the Fermi level E_F to shift downwards and crossing the enhanced density of states (DOS) shown in Figure 11D. Furthermore, some clear magnetic hysteresis loops at room temperature (RT) can be observed in FGT samples with different thicknesses, according to Polar-MOKE measurements (Figure 11C). The T_C of FGT can be increased to above RT through light control, providing many opportunities for the development of spintronic applications for 2D magnets.

4.5. Electrical Control

Previous studies have shown that electric fields modify the magnetism of metal films [159,160,161] and Fe/MgO junctions [162] by influencing the behavior of the electrons. Recently, Wang et al. [83] calculated the effect of the electric field on the magnetic anisotropy of the FGT monolayer, as shown in Figure 12A–C. The effect of orbital splitting caused by electron doping on magnetic anisotropy was more pronounced; meanwhile, the influence of hole doping related to orbital occupation was relatively weak. In addition, the change in magnetic anisotropy (Figure 12C) was more obvious in the single-gate configuration (Figure 12B).

Additionally, the generation of negative differential conductance (NDC) [84] can also be driven by a local electric field in FGT (Figure 12D). Furthermore, the three peaks in the Fe d orbits underwent significant shifts under the electric field. As the electric field was enhanced, the off-plane FM of FGT weakened, resulting in a decrease in MAE. Remarkably, in single-layer FGT, the electric field induces charge transfer in the FGT monolayer in the field direction. Therefore, applying an electric field has become an effective way to mediate 2D FM.

4.6. Proximity Effects

Proximity effects [85,163,164,165,166,167] are another dominant area of the research into 2D materials. For example, by using 2D magnetic materials adjacent to a bulk semiconductor substrate [168] or 2D materials with strong spin–orbit coupling [169], their magnetism can be enhanced. Intriguingly, Zhang et al. [85] fabricated antiferromagnetic FePS₃(FPS)/ferromagnetic Fe₃GeTe₂(FGT) heterostructures and detected the enhancement of T_C and H_C through proximity coupling effects. Furthermore, FPS/FGT/FPS exhibits a slightly different modulation of H_C compared to FPS/FGT, which is related to AFM-FM coupling. Moreover, the long-range magnetic order induced by topology triggered by femtosecond laser pulses [57] could also be maintained at room temperature.

The aforementioned methods for mediating magnetism require additional equipment, such as a three-point-bending experimental setup [48,72], DAC, [78], femtosecond laser excitations [53], and the scanning tunneling microscopy (STM) tip [84]. Directly exfoliating FGT nanosheets [88] onto different substrates could also tailor magnetism, as shown in Figure 13. Intriguingly, FGT samples with different thicknesses all exhibited magnetism, while samples on different substrates exhibited different T_C, indicating that the substrate has a modulation effect on T_C values. Furthermore, the lattice distortion and charge redistribution at the interface were related to substrate-induced FM, and this mechanism requires further exploration. Actually, substrates did not only affect the growth of 2D materials, but also determined their performance [170,171].

4.7. Doping Engineering

4.7.1. Doping with 3d Transition-Metals

Doping 3d-transition metal atoms is an effective strategy for controlling magnetism [66,69,90,172,173,174,175,176]. Theoretical calculations have shown that almost all 3d-transition metal atoms (except for Co atoms) [97] are more inclined to replace Fe1 atoms (Figure 2A). The charge transfer generated by doping atoms weakens the magnetic moment of Fe atoms, while the weakening effect of Fe1 atomic magnetic moment is more significant. However, the magnetism increases after doping with Co atoms, which may be related to the shrinking of the a-axis lattice constant. In experiments, doping 3d-transition metal atoms in bulk single-crystal samples were usually achieved via CVT [93] or self-flux [38]. Doping Ni atoms suppressed the ferromagnetic order, which rapidly decreased with the increase in the doping amount. The T_C decreased from 212 K to 50 K, and after reaching 0.44, the magnetic moment remained almost constant. Furthermore, the long-range magnetic order was suppressed and subsequently transformed into a glassy magnetic phase (Figure 14). However, doping Co atoms may cause an increase in H_C and the appearance of hard magnetic phases; this is related to the movement of pinned domain walls [8].

Bulk F5GT single crystals were also doped with Co atoms via CVT [95,96], as shown in Figure 15. As the amount of Co used for doping increases, it can drive the evolution of the lattice and of magnetism (Figure 15B). However, the nominal doping concentration was slightly different from the measured one, with only a specific concentration being more consistent. Afterward, Co atoms were doped into the lattice, resulting in a slight increase in their interlayer spacing. Indeed, a phase transition from FM to AFM occurred at high doses of doping in Figure 15C. However, Tian et al. [96] found that doping with 20% Co could increase its T_C to 337 K and induce complex magnetic phase transitions at higher Co doping levels. Furthermore, hexagonal 2D Co_yFe_5−yGeTe₂ (Figure 15E) and Ni_yFe_5−yGeTe₂ nanoflakes (Figure 15E) were prepared via flux-assisted growth (Figure 15D) [44]. Various elements in the nanosheets were evenly distributed via energy dispersive spectroscopy (EDS) elements mapping, and there were significant differences in the energy spectra of samples doped with different amounts of Co, with a doping amount of up to 66.7%. Furthermore, the doping of Co atoms caused a decrease in T_C, and as the doping amount increased, its T_C decreased even further (Figure 15F). In addition, the magnetic anisotropy underwent significant changes. Similarly, Ni doping can also cause a decrease in T_C. In other words, the higher the content of Fe, the higher its T_C.

4.7.2. Doping with Non-Metallic Atoms

Not only can Fe atoms be substituted with Co or Ni atoms [38,44,93,95,96,97], but doping can also be achieved by replacing Ge atoms with As atoms [92,98]. The doping of As atoms caused a decrease in the a-axis lattice constant and an increase in the c-axis lattice constant, thereby reducing the density of spin states below the Fermi level, resulting in a decrease in T_C [92]. Furthermore, its M_S decreased linearly with the increase in the doping amount in polycrystalline Fe_3−yGe_1−xAs_xTe₂ (0 ≤ x ≤ 0.85). Similarly, the expansion of the F5GT unit cell [98] in the c-axis direction and the contraction in the ab plane was also observed after doping with the As atom. In addition, its T_C and M_S decreased in polycrystalline Fe₅Ge_1−yAs_yTe₂ (0 ≤ y ≤ 1), a phenomenon similar to that observed in the Fe_3-yGe_1−xAs_xTe₂ (0 ≤ x ≤ 0.85) samples. Moreover, the stacking disorder caused by the local AFM coupling can reduce its M_S.

4.7.3. Electron Doping

Remarkably, Deng et al. [14] found that FGT devices could be operating in ionic gates, which provides a new approach for mediating 2D FM. Although they did not fully explain the relationship between it’s ferromagnetism and electron doping, the importance of this strategy was acknowledged. Soon after, gate-control was implemented to regulate magnetic resistance [99], magnetic phases [62,101], and interlayer coupling [100,102,177]. Furthermore, the T_C and H_C in FGT flakes [101] were decreased after Li⁺ doping from lithium-ion-conducting glass-ceramics (LICGC). In addition, electron doping influenced the Fe–Ge plane in the middle of the FGT monolayer, weakening it’s resistance and enhancing it’s T_C [105]. As a typical example, the modulation of interlayer coupling was achieved by fabricating FGT hall devices on solid-state proton conductors (Figure 16A). Furthermore, they discovered a clear exchange bias (EB) when changing the gate voltage (Figure 16B), which may be related to the presence of an AFM phase at low temperatures (Figure 16C). However, a random exchange bias (Figure 16D) occurs after applying a higher gate voltage. Furthermore, the exchange bias and coercivity undergo a complex evolution with the measurement times, but there have also been cases of small EB values and large coercivity. In addition, the type of AFM–FM interface coupling determines the positive and negative exchange bias (Figure 16E).

Additionally, Tan et al. [62] achieved the high-electron-concentration doping of F5GT through a solid proton conductor (Figure 17A,B). When a positive bias voltage was applied, the transport properties were unchanged. When a negative bias voltage was applied, there was a significant change in the transport properties caused by proton intercalation, especially when it reached −5 V, and the anomalous Hall loop disappeared (Figure 17C), accompanied by the appearance of a magnetic phase transition (Figure 17D,E). Moreover, different calculation methods have revealed that electron doping achieves a reversal of Hall conductivity and phase transition (Figure 17E). Therefore, electron doping or protonic gating indeed represents an efficient method of controlling magnetic phase transitions. Furthermore, Tang et al. [102] found that magnetic anisotropy in F5GT was continuously mediated by electrolyte gating. Moreover, the screening effect of itinerant electrons drives magnetic anisotropy to switch from an off-plane easy axis to an in-plane easy axis.

4.7.4. Hole Doping

Inspired by the gate-mediated RTFM in FGT thin flakes [14], many attempts have been made to control its ferromagnetism through hole [43,94,106] or electron [43,94,105] doping. In particular, the magnetic anisotropy in exfoliated Fe_2.75GeTe₂ flakes (Figure 18A,B) was inhibited by hole doping, resulting in a decrease in H_C (Figure 18C). The magnetic anisotropy could undergo a 93% attenuation, but the change in the magnetic moment was very small, as shown in Figure 18D. Furthermore, the electronic structure of Fe_2.75GeTe₂ single crystals changes due to hole doping, causing significant changes in magnetic anisotropy. In addition, another report [94] suggested that hole doping was beneficial for maintaining the long-range ferromagnetic order.

Remarkably, the intrinsic Fe vacancies [106] were probed via STM, as presented in Figure 19A–D. The peak near 20 mV originated from the Kondo lattice [106,121], and this Fe vacancy is called a Kondo hole (Figure 19C,D). Hole doping elevated the energy band, and the Fermi surface of FGT was shifted towards a lower energy level. After the formation of Fe vacancies at the Fe (II) site, the magnetic moment near the Fe (I) site decreased (Figure 19E,F), accompanied by the appearance of a higher charge density. Furthermore, the introduction of Fe vacancies reduced the magnetic moment near them, further strengthened the Kondo screening effect, and, thus, weakened the magnetism, as detailed in Figure 19G,H. The Kondo holes can affect the charge distribution of their own sites, and converting them into momentum space has a more significant impact. In other words, hole doping weakened the ferromagnetism of FGT.

4.8. Intercalation or Irradiation

Recently, inserting sodium into Fe_2.78GeTe₂ powders [107] can raise its T_C to ~300 K, as shown in Figure 20. After intercalating Na, more exposed edges appeared, and their layered features remained unchanged in a single crystal structure (Figure 20A,B). More specifically, the Fe, Ge, and Te elements (Figure 20C) were evenly distributed in the sample, while the inserted Na was concentrated at the edge. A phase transition occurred from PM (Fe_2.78GeTe₂) to FM (NaFe_2.78GeTe₂) at 200 K (Figure 20D), and the M_S was enhanced (Figure 20E). Furthermore, the magnetic hysteresis loops are also measured at 350 K. Notably, impurity phases, such as Fe or Fe_2−xGe, dominated the RTFM in the NaFe_2.78GeTe₂ samples. Alternatively, the T_C and exchange bias could be mediated with Fe-intercalation [108], which induces magnetic order by reinforcing magnetic coupling. However, the detected Te_Ge antisite defects had no modulation effect on the T_C of different samples. Thus, Na intercalation provides a novel strategy for enhancing T_C, which is related to the tensile strain.

Inspired by pattern-induced ferromagnetism [16], Yang et al. [109] also improved the T_C of FGT to 450 K via Ga irradiation. Irradiation induces an amorphous structure on the surface of the sample, which facilitates the formation of magnetic vortex states in FGT micropatterns. Moreover, irradiation causes a magnetization transition.

4.9. Twisting

Twisting 2D materials can introduce some novel properties, such as magnetism [178,179] and superconductivity [180], which trigger the interaction topology with magnetism in 2D ferromagnets, resulting in the formation of skyrmions [181,182] or magnons [183,184] in the twisting system. In fact, the stacking order directly affects the magnetism of bilayer CrI₃ by changing the crystal structure [178] or interlayer magnetic coupling [178]. Surprisingly, the magnetism was obtained in double bilayer CrI₃ [185] at small twist angles. Although the phase transition from AFM to FM has been theoretically achieved in twist-stacking bilayer FGT [110,179], it has not yet been experimentally achieved [186].

4.10. Patterning

Magnetic domain patterns on FGT surfaces can be modulated with various mechanisms [8,13,187,188], one of which is the phase transition from FM to AFM related to interlayer coupling [13]. The photoemission electron microscopy (PEEM) image in Figure 21A–E clearly shows the magnetic domain structure of FGT nanosheets, and the stripe domain structure disappears after reaching the T_C of 230 K. After patterning the FGT sample into diamond and rectangular shapes using a focused ion beam (FIB), striped magnetic domain structures, similar to those in the unpatterned FGT (Figure 21B,C), were also observed, as shown in Figure 21D. However, the striped domain structure did not completely disappear and was significantly weakened at 230 K (Figure 21C).

In addition, there were only in-plane magnetic domains in the patterned FGT at 240 K (Figure 19D); they disappeared at ~370 K, indicating that the T_C of bulk FGT was increased to 370 K. Notably, a novel magnetic vortex state or magnetic multidomain state was developed in patterned FGT (Figure 19D) at 300 K. Furthermore, spin reorientation occurred with increasing temperatures (Figure 19E). When using FIB to construct FGT patterns, Ga ions were unintentionally implanted into the sample, which may have caused the increase in T_C; however, to date, this hypothesis remains speculative.

5. Band Structure of Ferromagnetic Fe_xGeTe₂

Like its bulk form, the FGT monolayer is metallic, as shown in Figure 22A–C. Its band structures near the Fermi level can mainly be attributed to the contribution of the Fe 3d orbitals. Moreover, it was confirmed that the FGT monolayer has the itinerant FM order according to Stoner’s criterion [46,189]. Remarkably, the Stoner model related to itinerant electrons can be used to better elucidate the spontaneous magnetization in most 2D metallic ferromagnets. In addition, the electronic band structures of all the Fe_xGeTe₂ systems are metallic (Figure 22D–G), similar to the FGT monolayer.

Furthermore, the F4GT and F5GT bilayer [190] have band structures similar to the FGT bulk and FGT monolayer, exhibiting metallic magnetic properties. However, there is a significant difference in the polarizability of F4GT and F5GT near the Fermi level, which leads to their different transport characteristics. The unique nature of FGT gives its related devices many advantages, including nonvolatility, low reversal magnetic field, and the magnetic field reversal Fe_xGeTe₂ electrodes through the spin-polarized current.

6. Fe_xGeTe₂-based Devices

To the best of our knowledge, four typical devices have been constructed based on metallic Fe_xGeTe₂ ferromagnets, including magnetic tunnel junctions (MTJ) [112,113,114,115,136], tunneling spin valves [18,99,116,117], nonlocal spin valves [118] and spin–orbit torque (SOT) devices [20,119,191]. As shown in Figure 23A,B, a nonlinear behavior originating from tunneling characteristics [115] was exhibited in the I–V curve. Furthermore, a typical spin-valve behavior was also identified in the hysteresis loops (Figure 23C,D). After applying a specific voltage, the spin-transfer torque (STT) generated by the current caused the bottom FGT electrode to switch, which was closely related to MAE.

Regarding another typical device, Wang et al. [18] observed tunneling spin-valve behavior (Figure 24A,B) in an FGT/hBN/FGT heterostructure (Figure 25A). Its TMR reached up to 160%. The spin of the tunneling electrons created a nonlinear bias-dependent I–V curve related to Figure 24C. As the bias voltage increased, TMR exhibited a highly significant attenuation, as shown in Figure 24D. The inelastic tunneling channel related to bias led to spin relaxation, which may suppress TMR signals.

After applying a voltage to FGT/Pt hybrid devices (Figure 25A) [20], a current was generated between FGT and Pt, forming spin–orbit torques in Pt. Furthermore, a hard magnetic loop (Figure 25B) similar to an FGT device has also been observed in FGT/Pt devices. As the applied in-plane magnetic field H_x was increased, its transition current decreased (Figure 25C–F), regardless of the direction of H_x. This switch was related to the magnetic domain and domain walls. Moreover, the low switching current of the FGT monolayer was beneficial for exploring more effective devices. Very recently, Wang et al. [191] increased the T_C of FGT to RT by the topological insulator Bi₂Te₃ and achieved the SOT-driven magnetization switching at RT.

More significantly, a nonlocal spin valve device [118] that can operate at RT has also been successfully constructed. In addition, highly efficient spin manipulations were observed at the F5GT/graphene interface.

7. Magnetic Skyrmions in Metallic Fe_xGeTe₂

As topological magnetic materials, the spin textures in Fe_xGeTe₂ are also regulated by the Fe stoichiometry. For instance, Bloch-type skyrmion bubbles [192] were observed in FGT by using Lorentz transmission electron microscopy (LTEM). However, in another study [193], Néel-type skyrmions were reported, which indicated the existence of Dzyaloshinskii-Moriya interaction (DMI) in FGT. Generally, FGT was considered as a centrosymmetric material in the space group P6₃/mmc [35], which should not possess the asymmetric DMI. A possible interpretation is that the asymmetric interface between the pristine and oxidized FGT may induce an interfacial DMI [193].

More recently, a comprehensive study was implemented to answer this question more clearly [194]. It is found that the crystal structure of Fe_3−xGeTe₂ can be tuned by the Fe stoichiometry, that is, Fe_3−xGeTe₂ changes from centrosymmetric P6₃/mmc space group to non-centrosymmetric P6₃mc space group when x is larger than ~0.2 [194]. Notably, the P6₃mc space group allows an in-plane DMI in Fe_2.74GeTe₂, leading to the formation of Néel-type skyrmions. Whereas in Fe_3.00GeTe₂, the centrosymmetric P6₃/mmc structure forbids the DMI. Thus, Bloch-type skyrmion bubbles were observed when the dipole–dipole interaction and magnetic anisotropy obtained a delicate balance.

Moreover, in the ab-plane of F5GT, Bloch-type skyrmion bubbles with unconventional helicity polarization were observed [195,196], which shows special application potential for data storage and processing using the helicity freedom of skyrmions in the future. Additionally, in the ac and bc-plane of F5GT, (anti)meron chains were observed [197]. The tunable topological spin textures pave a promising way for constructing novel spintronic devices using Fe_xGeTe₂.

8. Outlook

In this review, we introduced the developments in and structures of 2D metallic Fe_xGeTe₂ ferromagnets. Then, we summarized six experimental methods, ten FM modulation strategies, and four typical spintronic devices that use 2D Fe_xGeTe₂ materials. Finally, we outlined the challenges and potential research directions in this field.

The samples prepared using the SSR, CVT, and flux methods were mainly bulk single crystals. Since the successful exfoliation of mono- or few-layer FGT, 2D Fe_xGeTe₂ has only been obtained experimentally. Very recently, CVD and MBE have also been used to prepare these materials. However, the lateral dimension of samples obtained via CVD was small and it was very difficult to accurately control the number of layers. Although MBE can be used to prepare wafer-scale materials and achieve 2D mode growth, it requires a high vacuum environment and is costly. So far, there is still no cost-effective method for preparing wafer-scale materials with controllable layers. In addition, new experimental methods, such as substitution reactions [198], are also worth further exploration.

Although ten strategies for regulating ferromagnetism have been proposed, not all of them have been achieved experimentally. Among these methods, twisting was only achieved through theoretical regulation. It is thus necessary to develop more strategies that could be executed experimentally to obtain RFTM in 2D Fe_xGeTe₂ materials. Furthermore, it would also be highly valuable to obtain FGT materials with good air stability [199] via new methods, as this would provide more convenient conditions for fabricating devices [200].