---
title: 'Fe3GeTe2: A 2D van der Waals Ferromagnet'
url: https://www.emergentmind.com/topics/fe3gete2-fgt
type: topic
---

# Fe3GeTe2: A 2D van der Waals Ferromagnet

Fe\(_3\)GeTe\(_2\) (FGT) is a layered, metallic van der Waals ferromagnet that combines itinerant \(3d\) magnetism, strong perpendicular magnetic anisotropy, and pronounced Berry-curvature-driven transport. In bulk, few-layer, and monolayer forms, it has served as a model system for two-dimensional Ising ferromagnetism, anomalous Hall transport, spin-current generation, spin-orbit torque switching, exchange bias in all-van-der-Waals heterostructures, and field- or interface-stabilized domain textures including stripe domains and skyrmion bubbles [1803.02559, 1903.00571].

## 1. Crystal structure and electronic character

FGT is generally described as a hexagonal van der Waals material with space group \(P6_3/mmc\), built from Fe–Ge–Te layers stacked along the crystallographic \(c\)-axis and separated by weak van der Waals gaps. Across the cited literature, the layer-resolved description varies in emphasis: several studies describe Te–Fe–Ge–Fe–Te sandwich layers or Te–Fe\(_3\)Ge–Te triple layers, whereas one exchange-bias study describes Te–Fe–Ge–Fe–Fe–Ge–Fe–Te slabs stacked along \(c\). A monolayer step height of about \(0.8\) nm is repeatedly reported, and the material can be mechanically exfoliated to the monolayer limit [1803.02559, 1903.00571, 1803.02934, 1902.05794, 2306.02129].

FGT remains metallic in few-layer and monolayer-derived contexts, which distinguishes it from many insulating or semiconducting van der Waals magnets. First-principles calculations place Fe-\(d\) bands at the Fermi level, hybridized with Te-\(p\) states, and several works describe topological or nodal-line features accompanied by strong spin–orbit coupling and large Berry curvatures near \(E_F\). In monolayer FGT, the cited symmetry is \(P\bar{6}m2\), with threefold rotation, three vertical mirror planes, and a horizontal mirror \(M_z\); in bilayer FGT, \(P\bar{3}m1\) is quoted, with inversion symmetry replacing \(M_z\). Those thickness-dependent symmetry changes are central to the allowed anomalous Hall and spin-current tensors [2104.05308, 2403.15020, 2304.13309].

Epitaxial growth on topological insulators further illustrates the structural fidelity of thin FGT. In molecular-beam-epitaxy-grown FGT on Bi\(_2\)Te\(_3\), scanning tunneling microscopy resolved \(a_{\rm FGT}=390\pm10\) pm and \(a_{\rm Bi_2Te_3}=430\pm10\) pm, together with a moiré periodicity of \(4.3\pm0.4\) nm attributable to lattice mismatch at zero rotational misalignment. The same study identified complete-QL Te-terminated regions, double-QL regions, and partial terminations, particularly a FeGe-terminated surface with a \(\sqrt{3}\times\sqrt{3}R30^\circ\) reconstruction [2303.10113].

## 2. Magnetic order, perpendicular anisotropy, and dimensional crossover

FGT is an out-of-plane ferromagnet with strong perpendicular magnetic anisotropy (PMA), but its reported magnetic scales are strongly sample-, thickness-, and protocol-dependent. Reported bulk or bulk-like Curie temperatures include \(T_C \simeq 160\text{–}190\) K from anomalous-Hall disappearance near \(180\) K in CrSBr/FGT studies, \(T_C \simeq 206\) K in twisted FGT/FGT junctions, \(T_C \simeq 205\text{–}208\) K in nanoflake anomalous-Hall measurements, \(T_{C0}=225(5)\) K in neutron powder diffraction under ambient pressure, and \(T_C \simeq 224.5\) K in exfoliated FGT/Pt bilayers by Arrott-plot analysis. In thinner flakes, a Pt/FGT switching study reported \(T_c\approx158\) K, and monolayer FGT was reported at \(T_C \simeq 130\) K [2306.02129, 2008.08313, 1803.02934, 2504.01306, 1903.00571, 1902.05794, 1803.02559].

The thickness dependence reveals a dimensional crossover rather than a single universal magnetic scale. For thicknesses below about \(4\) nm, corresponding to roughly five layers, the critical behavior crosses over from 3D to 2D Ising-like. Reported critical exponents from remanent magneto-optical measurements are \(\beta\approx0.25\text{–}0.27\) for thick flakes, \(\beta\approx0.20\pm0.02\) for intermediate thicknesses, and \(\beta\approx0.14\pm0.02\) in the monolayer, consistent with the 2D Ising value \(0.125\). A separate finite-size-scaling analysis of nanoflakes gave a coupling length of \(N=4.96\pm0.72\) van der Waals layers and \(\lambda=1.66\pm0.18\), indicating interlayer magnetic coupling over roughly five layers [1803.02559, 1803.02934].

Reported anisotropy parameters also span a wide range, again reflecting different thicknesses and analysis schemes. Examples include \(K_u(2\,{\rm K})\simeq3.5\times10^5\ {\rm J/m^3}\) for a \(10.4\) nm nanoflake, \(K_u \simeq (1\text{–}1.5)\times10^6\ {\rm J/m^3}\) in micromagnetic parameterizations chosen to reproduce the “hard” magnetic character of FGT, \(K_u^{\rm FGT}\approx1.46\ {\rm MJ/m^3}\) at \(5\) K in CrSBr-coupled heterostructures, and \(K_u\approx(0.3\text{–}0.5)\ {\rm MJ/m^3}\) in Lorentz-TEM studies. The anisotropy field is commonly written as
\[
H_k=\frac{2K_u}{\mu_0 M_s},
\]
and several studies infer \(H_k\) of order \(1\) T or several tesla depending on the adopted \(M_s\) and thickness [1803.02934, 2306.02129, 2508.00082, 2105.00468].

At low temperature, FGT frequently shows square hysteresis loops and hard-magnetic behavior. Reported coercivities include \(552.1\) mT at \(2\) K for a \(10.4\) nm nanoflake, up to about \(1\) T at \(2\) K in atomically thin magneto-optical measurements, and a typical increase from \(\sim0.5\) kOe at \(200\) K to \(\sim3\) kOe at \(2\) K in individual FGT flakes. Thick flakes also support multidomain states: above a critical thickness of about \(15\) nm, an intermediate temperature regime develops labyrinthine domains with characteristic widths around \(200\) nm [1803.02934, 1803.02559, 2008.08313].

## 3. Berry curvature, anomalous Hall transport, and spin-current generation

The anomalous Hall effect (AHE) is one of the main diagnostics of FGT magnetism and one of its principal intrinsic transport phenomena. In transport form, the Hall response is often written as \(R_{xy}(B)\simeq R_0B+R_sM_z(B)\), while the intrinsic anomalous Hall conductivity is evaluated from the Berry-curvature Kubo formalism. One cited expression is
\[
\sigma_{AHE}=-\frac{e^2}{\hbar}\int_{BZ}\frac{d^2k}{(2\pi)^2}\,\Omega(\mathbf{k}),
\]
and the strain study gives the clean-limit three-dimensional form for \(\sigma_{\alpha\beta}^{AH}\) and \(\Omega_{n,\alpha\beta}(\mathbf{k})\) with a \(120\times120\times60\) \(k\)-mesh and a broadening \(\Gamma=0.0129\) eV. In that work, the intrinsic anomalous Hall conductivity at the true Fermi level was reported as \(232.49\ (\Omega\,{\rm cm})^{-1}\) at \(0\%\) strain, \(220.23\ (\Omega\,{\rm cm})^{-1}\) at \(2.5\%\) armchair strain, and \(244.71\ (\Omega\,{\rm cm})^{-1}\) at \(5\%\) armchair strain, leading to the conclusion that the intrinsic AHE is robust to in-plane uniaxial strain up to \(5\%\) [2304.13309, 1803.02934].

That robustness is not universally consistent with all prior experimental interpretations. The same strain analysis explicitly notes disagreement with an earlier experimental report that less than \(1\%\) uniaxial strain can double the anomalous Hall resistance. Three explanations are discussed there: strain-sensitive extrinsic AHE terms, correlation effects beyond the simple DFT-level description, and concurrent changes in \(\rho_{xx}\) that alter measured \(R_{AH}\) without comparably altering \(\sigma_{xy}\). The authors state that this discrepancy implies that the present understanding of the AHE in FGT is incomplete [2304.13309].

FGT also supports a spin anomalous Hall effect (SAHE) whose magnitude and spin polarization depend nonlinearly on the magnetization direction. For an in-plane electric field \(E\parallel y\), the cited definitions are
\[
\sigma_{AHE}\equiv J_x/E_y,\qquad \sigma_{SH,\gamma}\equiv J^s_{x,\gamma}/E_y,
\]
with \(\gamma=x,y,z\). In monolayer FGT, a horizontal mirror \(M_z\) forbids an out-of-plane spin current for strictly in-plane magnetization, but once the magnetization acquires a \(z\)-component, both in-plane and out-of-plane spin-current channels become allowed. In bilayer FGT, the loss of \(M_z\) and the presence of inversion symmetry permit all three spin-polarization channels, so an intermediate magnetization direction can produce an arbitrary spin-polarization vector. A central relation derived for the intra-band part is
\[
\sigma_{SH,\gamma}^{\rm intra}\simeq \eta(\hat M)\,\sigma_{AHE}\,m_\gamma,
\]
linking SAHE directly to AHE through Berry curvature weighted by the local spin expectation value [2104.05308].

## 4. Spin–orbit torques and current-driven switching

FGT has been studied in both interfacial-SOT and bulk-SOT regimes. In Pt/FGT bilayers, a charge current in Pt generates a transverse spin current via the spin Hall effect and applies damping-like and field-like torques to the FGT magnetization. The torque forms quoted in the literature include
\[
\mathbf{T}_{DL}=\tau_{DL}\,\mathbf{m}\times(\mathbf{m}\times\boldsymbol{\sigma}),\qquad
\mathbf{T}_{FL}=\tau_{FL}\,\mathbf{m}\times\boldsymbol{\sigma},
\]
and the damping-like efficiency is often written as
\[
\xi_{DL}=\frac{2e}{\hbar}\left(\mu_0 M_s t_{\rm FGT}\right)\left(\frac{\Delta H_L}{J_e}\right).
\]
In few-layer Pt/FGT Hall bars, full hysteretic switching at \(100\) K under \(\mu_0H_x=\pm50\) mT occurred for \(I_c\approx8\) mA, corresponding to \(J_c\approx5\times10^7\ {\rm A/cm^2}\) in Pt. Harmonic Hall analysis at \(90\) K yielded \(\Delta H_L/J=(53.4\pm4.7)\ {\rm mT}/(10^7\ {\rm A/cm^2})\) and \(\Delta H_T/J=(24.3\pm1.3)\ {\rm mT}/(10^7\ {\rm A/cm^2})\), with an inferred \(\xi_{DL}\sim0.12\text{–}0.15\) when a reduced few-layer magnetization is assumed [1902.05794].

A separate FGT/Pt study on \(15\text{–}23\) nm flakes reported switching at \(180\) K with zero in-plane-bias extrapolated critical current density \(J_c(0)=2.5\times10^{11}\ {\rm A\,m^{-2}}\) in Pt. Second-harmonic Hall measurements gave \(\xi_{DL}=0.11\pm0.01\) at \(I_0=2.2\) mA and \(\xi_{DL}=0.14\pm0.01\) at \(I_0=2.4\) mA, values described there as rivaling the best metallic ferromagnet/heavy-metal stacks. That work attributes the large efficiency to the atomically flat FGT/Pt interface [1903.00571].

FGT also exhibits bulk spin–orbit torques without a heavy-metal overlayer. In single FGT flakes, symmetry analysis identifies a current-induced effective field
\[
B_{\rm SOT}=\Gamma_0J_0\left[(\mathbf{m}\cdot\hat e_y)\hat e_x-(\mathbf{m}\cdot\hat e_x)\hat e_y\right],
\]
and harmonic Hall measurements on a \(35\) nm flake gave damping-like fields up to \(8\ {\rm mT}/(10^{11}\ {\rm A\,m^{-2}})\). Fitting to the bulk form yielded \(|\Gamma_0|\approx50\ {\rm mT}/(10^{11}\ {\rm A\,m^{-2}})\) at low temperature, showing that current-driven manipulation in FGT need not rely exclusively on interfacial spin Hall injection [2107.09420].

These torque mechanisms have been extended to device architectures. An all-van-der-Waals three-terminal SOT-MRAM based on top-FGT/h-BN/bottom-FGT used the top FGT as the free layer and the bottom FGT as the pinning layer. At \(2\) K, the device showed an initial TMR ratio of about \(33\%\), a stable working TMR of about \(6\%\), onset of nonvolatile switching at \(I_c\simeq3.7\) mA, and \(J_c\simeq1.5\times10^6\ {\rm A/cm^2}\). The write and read current paths were physically decoupled, with in-plane mA-order writing current and vertical \(\mu\)A-order read current [2403.15020].

## 5. Exchange bias, tilted magnetic states, and voltage-controlled switching

Exchange bias in FGT has been demonstrated in several van der Waals heterostructures and is notable because the coupled antiferromagnets can have orthogonal magnetic anisotropies. In CrSBr/FGT, the in-plane easy axis of CrSBr acts on the perpendicularly magnetized FGT through an in-plane exchange-bias field \(H_{EB}\). The effect appears only below the CrSBr Néel temperature \(T_N\simeq132\) K, rises to about \(0.15\) T at \(10\) K, and vanishes above \(T_N\). A CrSBr thickness greater than \(10\) nm is required for non-zero exchange bias at \(30\) K. In a macrospin picture, the FGT tilt under this orthogonal coupling is described by
\[
E_{\rm tot}(\theta)=-M_sH_{EB}\sin\theta+K_u\sin^2\theta,
\]
which gives
\[
\sin\theta_{\rm eq}=\frac{M_sH_{EB}}{2K_u}.
\]
Using \(H_{EB}\simeq0.1\) T, \(M_s\simeq4\times10^5\ {\rm A/m}\), and \(K_u\simeq10^6\ {\rm J/m^3}\), the cited estimate is \(\theta_{\rm eq}\simeq50^\circ\text{–}60^\circ\), consistent with the greater-than-\(50^\circ\) tilt inferred from Hall data. In CrSBr/FGT/Pt, the same in-plane exchange bias provides sufficient symmetry breaking for deterministic spin–orbit torque switching at zero applied magnetic field [2306.02129].

A later CrSBr/FGT study linked this exchange bias to correlated domain structures rather than to a simple rigid interfacial shift. In Au/FGT \((87\pm2\ {\rm nm})\)/CrSBr \((30\pm2\ {\rm nm})\)/h-BN stacks, anomalous Hall loops at \(10\) K gave \(H_{EB}=-47\pm5\) mT after a \(+2.5\) T preset and \(H_{EB}=+45\pm5\) mT after a \(-2.5\) T preset. The bias vanished near \(130\) K, close to \(T_N\simeq132\) K of CrSBr, and a pronounced training effect was observed: only the first sweep after preset showed asymmetric switching and nonzero \(H_{EB}\). Off-axis electron holography on a \(bc\)-plane lamella revealed stripe-like flux-closure domains in FGT with circular rotation of the magnetic induction in the \(bc\) plane, directly connecting asymmetric reversal to domain nucleation and annihilation [2508.00082].

FGT can also host exchange bias against an antiferromagnetic oxide derived from itself. In FGT/O-FGT/hBN, where the O-FGT is formed by natural oxidation of the FGT surface, field cooling by \(\pm6\) kOe from \(300\) K to low temperature produced an exchange-bias field \(H_E\approx1.4\) kOe at \(10\) K and a blocking temperature \(T_B\approx150\) K. The gate dependence was linear,
\[
H_E(V_g)=H_E(0)+\left(\partial H_E/\partial V_g\right)V_g,
\]
with \(\partial H_E/\partial V_g\approx80\ {\rm Oe/V}\). In that device, deterministic voltage-controlled magnetization reversal occurred near \(V_g\sim+6\) V for stepped holds of \(60\) s, and the switching energy was estimated as \(U\sim20\) pJ per bit [2411.18278].

## 6. van der Waals heterostructures, magnetoresistance, moiré interfaces, and interfacial Hall control

FGT supports several distinct magnetotransport phenomenologies depending on what is stacked against it. In twisted FGT/FGT homojunctions with \(\theta\simeq87^\circ\), the interface remains metallic: the resistance decreases monotonically on cooling, from \(R_j\simeq400\ \Omega\) at \(300\) K to about \(350\ \Omega\) at \(2\) K, and the \(I\)–\(V\) characteristics are linear up to \(\pm5\) mV. The magnetotransport shows a plateau-like magnetoresistance defined by
\[
PMR=\frac{R_{AP}-R_P}{R_P}\times100\%,
\]
arising from antiparallel switching of the two FGT layers. The PMR grows from about \(0.025\%\) at \(100\) K to about \(0.05\%\) at \(2\) K. The same study emphasizes that this is at least three orders of magnitude smaller than typical TMR, consistent with a clean metallic junction rather than a tunnel barrier [2008.08313].

In FGT/graphite/FGT trilayers, the reported response is qualitatively different from conventional two-state giant magnetoresistance. The devices show three distinct resistance states: a high-resistance antiparallel “OUT” configuration, an intermediate parallel state, and a low-resistance antiparallel “IN” configuration. At \(50\) K, the MR amplitude \(\Delta R/R_0\) reaches about \(0.4\text{–}1.5\%\) depending on device and is essentially independent of graphite thickness between \(3\) and \(11\) nm. The proposed mechanism is spin-momentum-locking-induced spin-polarized current at the graphite/FGT interfaces rather than standard spacer-mediated GMR [1904.10588].

FGT also forms structurally clean interfaces with topological insulators. In monolayer FGT grown on Bi\(_2\)Te\(_3\), the moiré wavelength satisfies
\[
L_{\rm moiré}=\frac{a_{\rm FGT}a_{\rm Bi_2Te_3}}{|a_{\rm Bi_2Te_3}-a_{\rm FGT}|}\approx4.2\ {\rm nm},
\]
in agreement with the measured \(4.3\pm0.4\) nm. The topological surface state of Bi\(_2\)Te\(_3\) remained visible through quasiparticle interference, and magnetic circular dichroism on thin FGT/Bi\(_2\)Te\(_3\) films showed a square hysteresis loop at \(140\) K with coercivity around \(500\) Oe and \(T_C\simeq190\) K [2303.10113].

A different interfacial control mode appears in WTe\(_2\)/FGT. Applying a current through \(T_d\)-WTe\(_2\) modulates the AHE of adjacent FGT, with a relative change in AHE conductivity exceeding \(180\%\). The effect is absent in pure FGT, weakens as the FGT becomes thicker, and peaks for bilayer WTe\(_2\), which the authors attribute to a Berry-curvature-dipole mechanism in WTe\(_2\) and an inverse magnetic proximity effect on FGT. For one device, \(\sigma_{xy}(0)\simeq970\ \Omega^{-1}{\rm m}^{-1}\) at zero modulation current, rises to about \(1800\ \Omega^{-1}{\rm m}^{-1}\) by \(I_{\rm mod}=+0.5\) mA, and is suppressed to nearly zero near \(I_{\rm mod}\approx-0.5\) mA [2601.03744].

## 7. Domain textures, skyrmion bubbles, and external tuning

FGT supports multiple nonuniform magnetic states, and those states depend strongly on thickness, interfaces, and field history. In thicker flakes, labyrinthine domains appear in an intermediate temperature window, with typical domain widths around \(200\) nm. In Lorentz-TEM studies of \((001)\) FGT thin plates, the zero-field-cooled ground state below about \(130\) K consists of stripe domains with period \(\lambda_{\rm stripe}\approx120\) nm. Under a perpendicular field at \(110\) K, the stripes begin to break into mixed fragments and isolated bubbles near \(\mu_0H\approx360\) Oe, form fully developed skyrmion bubbles by about \(680\) Oe, and collapse to a uniform state above \(920\) Oe. A field-cooling protocol with \(\mu_0H=600\) Oe generated a hexagonal lattice with bubble diameters of about \(100\text{–}150\) nm and lattice constant \(a_{\rm hex}\approx150\text{–}200\) nm [1803.02559, 1912.11228].

The topological charge of those bubbles is described by
\[
N_{sk}=\frac{1}{4\pi}\int \mathbf{m}\cdot(\partial_x\mathbf{m}\times\partial_y\mathbf{m})\,dx\,dy,
\]
and transport-of-intensity reconstructions in the cited work are consistent with Bloch-type chirality and \(|N_{sk}|=1\) per bubble. An important point is that the micromagnetic model used for these FGT skyrmion bubbles did **not** require a DMI term: the textures were modeled from exchange, uniaxial anisotropy, Zeeman coupling, and dipolar energy in a centrosymmetric material. By contrast, in Pt/oxidized-FGT/FGT/oxidized-FGT heterostructures, interfacial DMI favors Néel-type walls. Lorentz-TEM showed that under \(\mu_0H_x=15\) mT the modulation wavevector \(q\) is perpendicular to the in-plane field in the heterostructure, whereas in an FGT plate without heavy-metal capping, \(q\) is rotated by \(\phi\approx53^\circ\), giving fan-like modulations characteristic of Bloch-like twists [1912.11228, 2105.00468].

FGT is also unusually tunable by pressure and chemical modification. Under hydrostatic pressure up to \(5\) GPa, neutron diffraction found no structural phase transition but a monotonic suppression of ferromagnetism: \(T_C\) decreases from \(225(5)\) K at \(0\) GPa to \(175(5)\) K at \(4.9\) GPa, with a slope \(k=10.0(5)\ {\rm K\,GPa^{-1}}\). The Fe(1) ordered moment at \(80\) K decreases from \(1.50\ \mu_B\) to about \(1.15\ \mu_B\) by \(4.8\) GPa, and the interpretation is that shorter Fe–Fe distances strengthen direct antiferromagnetic exchange while changes in Fe–Te–Fe and Fe–Ge–Fe angles weaken ferromagnetic superexchange [2504.01306].

Chemical intercalation can shift the system in the opposite direction. Electrochemical insertion of tetrabutylammonium cations into Fe\(_{3-x}\)GeTe\(_2\) yields Fe\(_{3-x}\)GeTe\(_2\)(TBA)\(_y\) with \(y\approx1\) per formula unit, injects roughly one electron per formula unit, and stabilizes a room-temperature ferromagnetic phase with \(T_C^{\rm intercalated}\approx350\) K, compared with \(T_C^{\rm pristine}\approx154\) K in the same study. Raman spectra show \(2\text{–}3\ {\rm cm^{-1}}\) red shifts of the \(E_g^2\) and \(A_g^1\) modes after intercalation, while vdW-corrected ab initio calculations indicate increased density of states at \(E_F\), enhanced nearest-neighbor \(J_1\) by about \(20\%\), and reduced magnetocrystalline anisotropy energy. This suggests that FGT is not defined by a single fixed magnetic parameter set; rather, it is a tunable itinerant ferromagnet whose ordering temperature, anisotropy, domain morphology, and transport response can be substantially reconfigured by thickness, interface design, pressure, gating, and intercalation [2209.08342].

Source: https://www.emergentmind.com/topics/fe3gete2-fgt