---
title: 'Fe3GeTe2: Layered vdW Ferromagnet'
url: https://www.emergentmind.com/topics/fe3gete2
type: topic
---

# Fe3GeTe2: Layered vdW Ferromagnet

Fe$_3$GeTe$_2$ (FGT) is a layered van der Waals ferromagnet that has been investigated as an itinerant metallic magnet, a correlated $3d$-electron system with Kondo-lattice phenomenology, a platform for chiral spin textures and multiple Hall responses, and an electrode or active layer in spintronic heterostructures. Most studies describe crystalline FGT as a hexagonal layered compound with strong perpendicular magnetic anisotropy and ferromagnetic ordering in the $\sim 150$–$230$ K range, while sputtered amorphous or interfacially modified forms exhibit semiconducting transport or proximity-induced magnetic behavior not present in pristine bulk crystals. Across these realizations, FGT has been used to probe electronic correlation, Bloch-state-controlled tunneling, bulk spin-orbit torque, exchange bias, and skyrmion physics [1512.05847][1904.06098][1907.08382][2309.06204].

## 1. Crystal chemistry and electronic structure

In the crystalline limit, FGT is usually described as a layered hexagonal compound with space group $P6_3/mmc$, often also characterized as a CdI$_2$-type structure. The basic structural motif is a Te–Fe–Ge–Fe–Te slab, with weak van der Waals coupling between adjacent slabs along the $c$ axis. Several studies emphasize two inequivalent Fe sites within the layer, a monolayer thickness of about $0.8$ nm, and honeycomb-like or slightly buckled Fe-based in-plane networks. One exchange-bias study instead describes FGT as rhombohedral with space group $R\bar{3}m$, so the crystallographic description is not entirely uniform across the literature summarized here [2112.13570][2008.10842][1907.08382][2303.13167].

Its low-energy electronic structure is consistently Fe-$3d$ dominated with substantial hybridization to Ge-$p$ and Te-$p$ states. Density-functional calculations for bulk FGT show a strongly spin-asymmetric Fermiology: majority-spin bands cross $E_F$ over large regions of the Brillouin zone, whereas minority-spin states form only a few small pockets, producing a density-of-states spin polarization near $E_F$ on the order of $70$–$80\%$. In transport calculations this asymmetry is central because the number of conducting channels at fixed $k_\parallel$ is much larger for the majority spin than for the minority spin. Beyond one-electron DFT, LDA+DMFT yields substantial narrowing of the $d$ bands in the $[-2\,\mathrm{eV},+2\,\mathrm{eV}]$ window, a coherence peak at $E_F$, and spectral density closer to photoemission than spin-polarized GGA alone, establishing the role of dynamical correlation in the ferromagnetic metal [1904.06098][1512.05847].

## 2. Magnetic order, anisotropy, and correlation effects

FGT is reported as a ferromagnet with a strong out-of-plane easy axis or perpendicular magnetic anisotropy, but the exact ordering scale depends strongly on sample thickness, stoichiometry, and device context. Bulk and thick-flake values are variously reported as $T_C \simeq 200$ K, $200$–$220$ K, $\sim 220$–$230$ K, and $233.5$ K, whereas one flux-grown single-crystal study reports $T_c \simeq 150$ K. For exfoliated nanoflakes, representative values include $185$ K for $21.3$ nm, $180$ K for $16.7$ nm, and $145$ K for $6$ nm, while monolayer FGT is reported near $130$ K. Representative magnetic parameters likewise vary by method and geometry: $M_s \simeq 3.25\,\mu_B/\mathrm{f.u.}$ for $H\parallel c$ at $5$ K in one single-crystal study, $M_s \approx 376\,\mathrm{emu\,cm^{-3}}$ at $5$ K in another, and a bulk moment of $1.58\,\mu_B/\mathrm{Fe}$ from magnetization and XMCD analysis. The uniaxial anisotropy is large, with values such as $K_u \simeq 1.46\times 10^6\,\mathrm{J\,m^{-3}}$ at $5$ K and $K_u(94\,\mathrm{K}) \simeq 9.7\times 10^5\,\mathrm{J\,m^{-3}}$ reported for thin heterostructure samples [1907.08382][2108.12111][2507.07458][1907.11349][1907.02397].

Electronic correlation is unusually prominent for a $3d$ ferromagnet. The experimentally determined Sommerfeld coefficient $\gamma_{\exp}=110\,\mathrm{mJ\,mol^{-1}\,K^{-2}}$ exceeds the GGA-derived $\gamma_{\mathrm{LDA}}\approx 8.3\,\mathrm{mJ\,mol^{-1}\,K^{-2}}$, giving $m^\*/m_b\simeq 13.3$, while LDA+DMFT produces a lower-bound renormalization of about $2.6$. STS and ARPES studies summarized in the Kondo-hole work further identify broad peaks at $-180$ mV and $-50$ mV assigned to itinerant ferromagnetic $d$ bands, a narrow Kondo-resonance shoulder at $+20$ mV, and heavy-fermion quasiparticles below a coherence scale. Within that framework, FGT is modeled as a Kondo lattice coexisting with Ising ferromagnetism, with nearest-neighbor Fe(I)–Fe(I) Ising exchange $J_H\approx 9$ meV and Fe vacancies acting as Kondo holes that enhance the local Kondo resonance, with $T_K\sim 300$ K inferred from a Fano half-width of $\Delta\approx 26$ meV [1512.05847][2008.10842].

Lattice dynamics couple directly to the magnetic state. Raman measurements on few-layer FGT identify two $A_{1g}$ phonons and one doubly degenerate $E_{2g}$ phonon at room temperature, with the $E_{2g}$ mode reversing photon helicity and thereby exhibiting phonon chirality with pseudo-angular momentum. Below the Curie temperature, the phonon frequency and linewidth deviate strongly from a symmetric anharmonic decay model, and the spin-phonon coupling constant extracted from the $A_{1g}(1)$ mode is $\lambda\approx 1.36\,\mathrm{cm^{-1}}$. Two additional modes around $240\,\mathrm{cm^{-1}}$ emerge only at cryogenic temperature as the Raman susceptibility is enhanced in the magnetically ordered phase [1909.01598].

## 3. Charge transport, Hall response, and spin-dependent tunneling

Magnetotransport measurements on single crystals identify a clear electron-magnon contribution to the resistivity. Below about $43$ K, $\rho(T)$ is well described by $\rho(T)=\rho_0+A T^2$; from $43$ K up to $T_C\approx 210$ K the best fit is $\rho(T)=\rho_0+A T^2+B T$; and above $T_C$ the resistivity is linear in temperature. The same study reports positive non-saturating magnetoresistance at low temperature for $H\parallel ab$, nearly linear negative magnetoresistance at high fields below $T_C$ due to suppression of magnons, and a Khosla-Fischer form $\Delta\rho(H)=-b_1\ln[1+(b_2\mu_0H)^2]$ above $210$ K, with $b_1\approx 1.08\,\mu\Omega\,$cm and $b_2\approx 0.39\,\mathrm{m^2\,V^{-1}\,s^{-1}}$. In the same data set the anomalous Hall coefficient follows $R_s=a\rho_{xx}^2+b\rho_{xx}$ with $|a|\gg|b|$, and the temperature dependence of the side-jump term tracks the field-induced change in resistivity attributed to magnon suppression. In that interpretation, the anomalous Hall effect is dominated by extrinsic side-jump scattering from magnons rather than by a purely intrinsic topological-band-structure contribution [2209.03555].

FGT is also notable as an electrode material in coherent van der Waals magnetic tunnel junctions. First-principles transport calculations for Fe$_3$GeTe$_2|$graphene$|$Fe$_3$GeTe$_2$ and Fe$_3$GeTe$_2|h$-BN$|$Fe$_3$GeTe$_2$ predict giant tunneling magnetoresistance,
$$
\mathrm{TMR}=\frac{G_P-G_{AP}}{G_{AP}},
$$
with $T_P^\uparrow+T_P^\downarrow=2.60\times10^{-2}$ and $T_{AP}=6.98\times10^{-4}$ for the graphene spacer, corresponding to $\mathrm{TMR}\simeq 3621\%$, and $T_P^\uparrow+T_P^\downarrow=2.15\times10^{-2}$ and $T_{AP}=3.38\times10^{-4}$ for the $h$-BN spacer, corresponding to $\mathrm{TMR}\simeq 6256\%$. The central mechanism is a Bloch-state mismatch in the antiparallel configuration: because spin and $k_\parallel$ are conserved in coherent tunneling, majority states on one side must connect to minority states on the other, but those states occupy disjoint regions of the two-dimensional Brillouin zone. The same work finds that replacing graphene or $h$-BN by vacuum preserves the qualitative $k_\parallel$-resolved transmission pattern, indicating that the giant TMR is set primarily by the bulk spin structure of FGT rather than by interface chemistry [1904.06098].

## 4. Chiral spin textures and topological transport

Direct imaging has established that thin FGT can host chiral textures without an added heavy-metal underlayer. Off-axis electron holography and Lorentz TEM on $20$ nm nanoflakes reveal zero-field-stabilized Néel-type skyrmions at cryogenic temperature, together with spirals and field-polarized ferromagnetic states in a temperature-field phase diagram. In zero field, a fluctuating spiral phase appears just below $\sim 150$ K and condenses into skyrmions that remain stable down to $10$ K; on warming, skyrmions persist to about $170$ K and then dissolve into spirals. Thickness is decisive: $15$–$25$ nm flakes show zero-field skyrmions, $25$–$35$ nm flakes show coexisting short spirals and skyrmions, flakes thicker than $35$ nm are spiral-dominated, and flakes thinner than $\sim 15$ nm show no domain contrast at zero field. From the thickness-driven skyrmion–spiral boundary the interfacial DMI is estimated as $D_i\simeq 0.94(\pm 0.17)\,\mathrm{mJ\,m^{-2}}$, while DFT+SOC gives $D_i\simeq 1.45\,\mathrm{mJ\,m^{-2}}$ [1907.08382].

Topological Hall signatures in FGT occur in several distinct geometries. A planar topological Hall effect was reported in a uniaxial single crystal when the magnetic field rotates into the hard $ab$ plane, with the PTHE persisting for all $T<T_c$ and reaching $\rho_{xy}^{PT}=2.04\,\mu\Omega\,$cm at $100$ K and $\mu_0H\approx 1.2$ T. That work attributes the signal to an internal gauge field produced either by topological domains or by a non-coplanar spin structure formed during in-plane magnetization. A later planar-geometry study found that below $T_C=233.5$ K a topological cusp in the Hall response coexists with positive longitudinal magnetoresistance dominated by spin-flip scattering, both confined to the unsaturated regime below a characteristic field $B^\*$ and to a narrow polar-angle window $\theta=90^\circ\pm 5^\circ$. In that data set, $B^\*(0)\approx 6$ T, the positive MR reaches about $+3$–$4\%$ at $5$ K, and the topological Hall amplitude is $\sim 0.1$–$0.3\,\mu\Omega\,$cm at the same temperature, supporting a common origin in a field-tilted, non-coplanar spin phase [1907.02397][2507.07458].

Interfacial and moiré engineering extend this topological sector. In WTe$_2$/FGT heterostructures, inversion-symmetry breaking and strong spin-orbit coupling from 1T$'$-WTe$_2$ induce a DMI of about $1.0\,\mathrm{mJ\,m^{-2}}$, yielding a topological Hall signal below $100$ K and directly imaged Néel-type skyrmions with diameters of about $150$ nm at $94$ K and $80$ nm at $198$ K; the zero-field stripe width is about $290$ nm. In twisted metallic FGT bilayers, a giant topological Hall effect appears only in the “magic” twist-angle window $0.45^\circ$–$0.75^\circ$, where $\rho_{xy}^{THE}$ reaches about $0.7\,\mu\Omega\,$cm between $20$ and $70$ K. Using
$$
\rho_{xy}^{THE}=R_0 P \Phi_0 n_{sk},
$$
that work extracts $n_{sk}\simeq 2.15\times 10^{15}\,\mathrm{m^{-2}}$ at $\theta=0.45^\circ$ and $60$ K, corresponding to an experimental skyrmion diameter of about $23$ nm, while micromagnetic simulation gives about $35$ nm and attributes the phase to alternating in-plane and layer-contrasting DMI with $D(0.6^\circ)\sim 4$–$6$ meV [1907.11349][2603.14192].

## 5. Disorder, proximity effects, exchange bias, and chemical substitution

FGT remains ferromagnetic even when long-range crystalline order is removed. Room-temperature magnetron sputtering produces amorphous films with no detectable FGT lattice fringes in HRTEM or STEM-EDS and no FGT reflections in XRD, yet XPS indicates that the local Fe, Ge, and Te valence states remain essentially identical to those of crystalline FGT. In these amorphous films, Arrott-plot analysis gives $T_C\approx 225$ K for thicknesses $\geq 5$ nm and $T_C\approx 100$ K at $1$ nm, closely tracking the thickness dependence reported for crystalline FGT. Magnetoresistance measurements further show two switching dips associated with domain-wall transport even in a single amorphous monolayer, leading that study to argue that long-range ferromagnetic order in FGT need not correlate with two-dimensional crystalline order [2302.00553].

A different sputtered realization is semiconducting rather than metallic. In sputtered Fe$_3$GeTe$_2$ films with atomic percentages Fe:Ge:Te = $42.1\%:24.6\%:33.3\%$, XRD shows no Fe$_3$GeTe$_2$ reflections even at $120$ nm, and all single layers from $3$ to $120$ nm show semiconducting $R(T)$ with $dR/dT<0$. Pure amorphous FGT exhibits an anomalous Hall effect only below about $180$ K, but FGT$(10\,\mathrm{nm})$/Pt$(3\,\mathrm{nm})$ shows square anomalous-Hall loops already at $300$ K with $H_c\approx 0.08$ T and $\Delta R_{AHE}\approx 0.4\,\Omega$, and VSM loops remain open at $400$ K. In that system, first-principles calculations for Pt adsorbed on a crystalline FGT monolayer give an enhanced interfacial exchange $J_{\mathrm{top}}\approx 15$ meV compared with about $8$ meV in bare FGT, while the measured spin Hall magnetoresistance ratio is about $0.12\%$ at $300$ K, with an effective spin Hall angle $\theta_{SH}\approx 0.097$ and $\lambda_{Pt}\approx 1.0$ nm [2309.06204].

Exchange bias and coercivity can also be engineered chemically. In Co-phthalocyanine/FGT heterostructures, the exchange-bias field reaches about $-840$ Oe at $10$ K for a $20$ nm FGT flake and remains measurable to approximately $110$ K, consistent with a spinterface picture in which CoPc spins acquire antiferromagnetic order by proximity and pin the FGT reversal. In CrPS$_4$/(O-FGT)/FGT heterostructures, the bias is smaller for a pristine interface, $\mu_0H_{EB}(5\,\mathrm{K})\approx 12.5$ mT, but reaches about $75$ mT for the naturally oxidized interface and shows non-monotonic temperature dependence with blocking scales near $20$ K, $70$ K, and $140$ K attributed to CrPS$_4$, FeO, and Fe$_3$O$_4$, respectively. Substitution on the Fe site provides another control parameter: in $(\mathrm{Fe}_{1-x}\mathrm{Co}_x)_3\mathrm{GeTe}_2$, increasing $x$ suppresses both $\theta_{CW}$ and the ordered moment, but near $x\approx 0.5$ the domain-wall pinning is strongly enhanced, with $H_c\approx 1.5$ kOe and $M_R/M_S>0.8$ for $0.46\le x\le 0.58$. That Co-doping study also argues that the low-field zero-field-cooled “kink” previously interpreted as an antiferromagnetic transition is better understood as a domain-wall pinning–depinning crossover [2302.11071][2303.13167][1905.03458].

## 6. Local probes, current-driven control, and nanostructures

Quantum sensing with spin defects in $h$BN provides spatially resolved access to both static and fluctuating magnetism in FGT. In one wide-field imaging platform, an exfoliated FGT flake of thickness $t_p\simeq 86$ nm is capped by an $\sim 88$ nm $h$BN layer containing boron-vacancy spin defects generated by He$^+$ implantation, with the defect-to-FGT distance estimated at $\sim 50$ nm. ODMR and relaxometry mapping show a perpendicular magnetization $\mu_0M_s\simeq 1.3$ kG at $6$ K, gradual demagnetization up to $225$ K, and domain-wall motion with saturation by about $600$ G at $178$ K. The relaxation rate peaks near the Curie region at $T_{\mathrm{peak}}\simeq 189$ K with $\Gamma_{\max}\simeq 36$ kHz, from which the authors extract $\chi_\parallel=(1.5\pm 0.2)\times 10^{-2}\,\mathrm{emu\,cm^3\,Oe^{-1}}$ and a spin-diffusion constant $D=(1.7\pm 0.3)\times 10^{-5}\,\mathrm{m^2\,s^{-1}}$ [2112.13570].

Current-induced torques in FGT are unusual because the monolayer has noncentrosymmetric $D_{3h}$ symmetry even though the three-dimensional bulk is commonly described by $P6_3/mmc$. Harmonic Hall measurements on a $35$ nm flake identify dominant bulk spin-orbit torques with damping-like effective fields per current density exceeding $50\,\mathrm{mT}/10^{11}\,\mathrm{A\,m^{-2}}$ at $100$ K, while the field-like term ranges from about $-2$ to $+4\,\mathrm{mT}/10^{11}\,\mathrm{A\,m^{-2}}$. The sign reversal between $\Phi=0^\circ$ and $\Phi=90^\circ$ is consistent with the bulk-symmetry-allowed form of the torque, and the torque persists from $20$ K to near $T_C$ [2107.09420].

The same current responsiveness enables direct coercivity control and memory operation. In nm-thin Hall-bar devices, increasing in-plane current density produces a monotonic reduction of the coercive field: around $J\approx 7\,\mathrm{mA/\mu m}$, $H_c$ is reduced by about $50\%$, and by $J\approx 30\,\mathrm{mA/\mu m}$ the hard magnetic state is driven to a soft state with $H_c\to 0$. The extracted torque coefficient is $|\Gamma_0|\simeq (50\pm 15)\,\mathrm{Oe}/(\mathrm{mA/\mu m})$, and a memory demonstrator switches at $J_{\mathrm{switch}}\approx 5\,\mathrm{mA/\mu m}$ under a small out-of-plane bias field. Nanoribbon calculations extend this device picture to the edge-dominated limit: monolayer FGT ribbons show non-collinear edge magnetism with canting angles of about $20^\circ$–$30^\circ$ decaying over roughly $8$–$10$ Å, an orbital Hall conductivity near $E_F$ about $8.5$ times the spin Hall conductivity, and deterministic spin-orbit- or spin-transfer-torque switching for $j_c\sim 10^{11}$–$10^{12}\,\mathrm{A\,m^{-2}}$ on sub-$100$ ps timescales in atomistic simulations [2108.12111][2502.12356].

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