Introduction

With the discovery of graphene and its astonishing physical properties,1 a new class of two-dimensional (2D) materials, ascribed to the dimensionality effect and modulation in their band structures, has become the focus of intense research which ranges from 2D semiconductors, like black phosphorous,2 MoS2,3 and WSe2,4,5,6 to 2D superconductors, for example, NbSe2,7, 8 and FeSe.9,10,11 The 2D ferromagnetic materials (2D FMs),12 were predicted to have promising spintronic applications13,14,15,16 with stable storage, faster response and low-power dissipation.12, 17 To this end, two approaches are currently adopted involving the introduction of defects or adatoms into parent materials or the proximity effect with ferromagnetic materials,18 for instance, the ferromagnetism in NbSe2 through hydrogen absorbed on the surface,19 and the search for the intrinsic magnetic materials. However, the resulting ferromagnetic materials by the former method significantly suffer from the undesired stability and limited controllability. To date, the predicted possible categories of 2D intrinsic FMs are mainly made up of Fe3GeTe2 20, 21 and CrXTe3 (X = Si, Sn, Ge)22,23,24 with the calculated Curie temperature of 220 K (bulk),25, 26 90 K (monolayer),24 170 K (monolayer),22 and 130 K (monolayer).24

Experimentally, some progress has been made. By reducing the sample thickness,18 V5S8 changes from antiferromagnetic to ferromagnetic states. Through previous ferromagnetic resonance (FMR) study,17 the exfoliated CrGeTe3 exhibits a Curie temperature around 61 K with an uniaxial magnetic property which can be tuned by external magnetic field.27 Even down to bilayer and monolayer, CrGeTe3, and CrI3 are still in the ferromagnetic state with a Curie temperature of 30 27 and 45 K,28 respectively, which is much lower than their bulk counterparts. Recently, the reported field-effect transistors based on ferromagnetic semiconductor C3N quantum dots (QDs)29 show an on–off ratio around 1010 with the tunable bandgap ranging from visible to infrared light for different-size QDs. However, the experimental progress in producing large-scale 2D FM thin films remains an obstacle. So far, the ferromagnetic single crystals such as Fe3GeTe2 and CrGeTe3 were mainly developed by chemical vapor transport (CVT).17, 25 Due to the van der Waals interaction between layers, these materials can be easily exfoliated by the Scotch-tape method, but yet, the resultant nano-devices are largely restricted by the sample size.30 Thus, the successful growth of large-scale 2D FM thin films becomes a thrust theme as it not only boosts the rapid exploration of physics but also renders the possibility to investigate various heterojunctions.12 For instance, by coupling with topological insulators, it is possible to probe quantized anomalous Hall effect and magnetic monopoles.31, 32 With superconductors, Andreev reflection effect7, 33 and the super-exchange interaction34, 35 can be investigated through the proximity effect. Besides the exotic physics, the molecular beam epitaxy (MBE) is well established in growing highly uniform and crystalline quality films. For the 2D materials, the thickness can be well controlled through the in situ reflection high-energy electron diffraction (RHEED) oscillations. These outstanding superiorities make MBE a desirable technique for producing wafer-scale-size films with well-defined thickness controllability.

Compared with other ferromagnetic materials such as CrSiTe3 and CrGeTe3, Fe3GeTe2 possesses outstanding properties. First, its superior stability is confirmed by the non-imaginary frequencies in phonon spectrum calculations.20 Indeed, the oxidation process of Fe3GeTe2 in our experiments is slower in comparison with CrGeTe3. Second, it is predicted to be an itinerant ferromagnet with a relatively high-Curie temperature around 220 K.26 Essentially, this becomes a very important figure of merit for spin-based technology architecture. Third, the anisotropy energy of Fe3GeTe2 around 107 erg/cm3 determined by magnetic force microscope36 is much larger than that of CrGeTe3 with 105 erg/cm3 measured with FMR,17 which is critical for magnetic tunnel junctions and magnetic random access memory devices. All of them require a large difference between anisotropy energy and thermal activation energy k B T for stable storage.20 In addition, the perpendicular magnetic anisotropy provides lower switching current in comparison with the in-plane one.37, 38

Here, we investigate 2D FM Fe3GeTe2 thin films grown on sapphire and GaAs by MBE. Periodic in situ RHEED oscillations clearly demonstrate the 2D growth mode through which the thickness of Fe3GeTe2 films can be precisely determined. Our low-temperature transport experiments unveil the out-of-plane easy axis along c direction with the Curie temperature of 216.4 K. And importantly, the Curie temperature is found to vary systematically with the Fe composition. By constructing a superlattice-like structure of (Fe3GeTe2/MnTe)3, where the subscript 3 means three periods of Fe3GeTe2/MnTe, the coercive field of Fe3GeTe2 increases 50%, exhibiting a tunable magnetic states.

Results and discussion

Two-inch high-crystalline-quality Fe3GeTe2 films were grown in a high-vacuum MBE system. Figure 1a shows the crystal structure geometry, which belongs to a space group P63/mmc. Each layer consists of five sublayers with sandwich-like stacking order. The lattice constants are a = b = 3.9536(7) Å and c = 16.396(2) Å. The top view displays a hexagonal distribution of atoms, as shown in the top panel of Fig. 1a. The inset of Fig. 1b exhibits a typical RHEED pattern obtained during the growth, where the streaky stripes suggest the atomically smooth surface and indicate the high-crystalline quality. These properties are further demonstrated by X-ray diffraction (XRD) and high-resolution transmission electron microscope (HRTEM). Figure 1b shows in situ RHEED oscillations and the equal periodicity provides a strong evidence of 2D growth mode. We note that it typically takes 111 s to complete one layer growth, suggesting that the thickness of Fe3GeTe2 films can be well controlled. Due to the low XRD intensity of 8 nm film, a thicker sample on sapphire under the same growth condition is used for the XRD measurements, as shown in Fig. 1c. According to PDF Card #75-5620 and the published XRD data of single crystals,26 peaks marked in red can be ascribed to a series of {002} planes—(002), (004), (008), (0010), and (0014), suggesting that Fe3GeTe2 is single crystalline and the growth is along c axis. Other peaks marked in blue are originated from the sapphire substrate. HRTEM experiments were further carried out to determine the layered structure of the Fe3GeTe2 film on (111) GaAs (Fig. 1d). The upper inset of Fig. 1d is the HRTEM taken from the Fe3GeTe2 region in which a van der Waals gap of 0.82 nm can be observed, corresponding to the (0002) displacing of Fe3GeTe2.20 Figure 1d shows the HRTEM image taken from the interface region between Fe3GeTe2 film and GaAs substrate, taken from GaAs {110} axis. The corresponding selected area electron diffraction (SAED) taken from the interface region is presented in Fig. 1e, from which the crystallographic relationship between the Fe3GeTe2 film and GaAs can be determined as {0001}Fe3GeTe2//{111}GaAs. Energy-dispersive X-ray spectroscopy (EDS) maps taken from the film region confirms the uniform elemental distribution of Fe, Ge, and Te in the film, as shown in Fig. 1f. With all the above characterizations, the Fe3GeTe2 thin films are found to grow uniformly on the substrate, and have a high-crystalline quality.

Fig. 1
figure 1

Fe3GeTe2 thin film growth and characterizations. a Structure geometry of Fe3GeTe2, terminated with Te layer. The top panel is a top view, where the hexagonal distribution can be observed clearly. b RHEED oscillations. By the periodic oscillations, the 2D epitaxial mode can be verified and the growth period per layer is 111 s. The inset is the RHEED picture with streaky stripes, demonstrating the smooth surface condition. c XRD data, compared with the PDF Card #75-5620 and the reported single crystals, the orientation can be determined to be along [002] directions, marked in red. d Cross-section HRTEM images taken from the interface area of the Fe3GeTe2 thin film grown on (111) GaAs with the scale bar of 2 nm. The upper inset marked by a red rectangle is taken from the Fe3GeTe2 region, indicating its layer distance is 0.82 nm, corresponding to {0002}Fe3GeTe2. e Corresponding SAED. f EDS mapping results, displaying uniform distribution of Fe, Ge, and Te atoms in the film, marked as the region between the marked white lines. The scale bar is 100 nm. Note that the Pt layer is deposited during the TEM sample preparation by focused ion beam (FIB)

Full size image

To explore low-temperature electrical properties, two-inch Fe3GeTe2 films were cut into small pieces and fabricated into a Hall-bar structure (Fig. 2a inset). Figure 2a shows the temperature-dependent resistance (R-T) data of 8 nm Fe3GeTe2 on Al2O3. Consistent with the theoretical predictions,21 with decreasing temperature the resistance decreases, showing a metallic behavior. However, upon reaching ~ 40 K, the R-T curve exhibits an upward increase. The inset shows a fit to the Mott’s variable range hopping (VRH) model yielding ln(R xx ) ~ T  −1/3, where 1/3 is characteristic for 2D materials and it would be 1/4 for 3D materials, which means that at low temperatures the localized carriers hop to the lowest activation energy states other than the nearest-neighboring states. A similar transport behavior was also observed in other material systems.39,40,41

Fig. 2
figure 2

Transport properties of 8 nm Fe3GeTe2 thin film. a R-T curve, showing a metallic characteristic. The bottom inset is a fit to the variable range hopping model, ln(R xx ) ~ T  –1/3. The scale bar of the top inset Hall-bar structure is 1 mm. b Angle-dependent anomalous Hall data. The measurement geometry is displayed in the inset. With magnetic field changing from perpendicular (θ = 0o) to parallel (θ = 90o) to the sample surface, the coercive field increases largely, from which the easy axis can be determined to be out-of-plane. c Anomalous Hall data at different temperatures with the offset of 50 Ω. With rising temperature, the coercive field decreases successively. The hysteresis at 200 K is very small, the coercive field at which is displayed in 2 e. d Carrier density and mobility at different temperatures. e Temperature-dependent coercive fields, which are calculated from anomalous Hall results (Fig. 2c). The error bars were estimated to be around 0.01 Tesla in the whole temperature range

Full size image

To investigate the ferromagnetic properties, a magnetic field up to ±9 T is scanned back and forth. Generally, the Hall effect of ferromagnetic materials can be described by42 R xy  = R H B + R AH M, where R H stands for the normal Hall contribution, and R AH comes from the magnetization contribution. Subtracting the normal Hall slopes from the raw data yields the anomalous Hall signals that are also dependent on the angle between magnetic field (B) and the normal vector of the sample surface (Fig. 2b and inset). It is evident that the easy axis of Fe3GeTe2 is along c-axis (out-of-plane) as the coercive field (H c) increases simultaneously with the angle. When raising the temperature, H c decreases and will reach zero at the Curie temperature, as shown in Fig. 2c and e, respectively. Based on that, the Curie temperature can be roughly estimated. More precise determination of Curie temperature can be achieved by the Arrott-plot, as described later. At 2 K, the hole carrier density and mobility were extracted to be 1.2 × 1019 cm−3 and 54.9 cm2 V− 1 s− 1, respectively, as displayed in Fig. 2d.

To estimate the Curie temperature and compare it with that of the bulk crystal counterpart, the method of the Arrott-plot has been adopted.41 First (R xy /R xx 2)2 is plotted against B/(R xy /R xx 2). It is well known that the intercept is positive for ferromagnetic and negative for paramagnetic state.43 Then, the Curie temperature T c can be extracted when the intercept on the (R xy /R xx 2)2 axis goes to zero. Figure 3a shows (R xy /R xx 2)2 ~  B/(R xy /R xx 2) curves of 70, 100, 125, 150, 180, and 200 K, and the intercept, determined from these linear fittings (Fig. 3b), decreases when the temperature increases. Using a liner fit, the Curie temperature is determined to be 216.4 ± 0.4 K.

Fig. 3
figure 3

Arrott-plots using anomalous Hall data and magnetization measured by SQUID. a and b are Arrot-plots at temperatures from 70 K to 200 K. The x-axis and y-axis are B/(R xy /R xx 2) and (R xy /R xx 2)2, respectively. Dashed lines represent the linear fits to the small-region data that near magnetization saturation. Using the intercepts in y-axis (summarized in Fig. 3b) and fitting these values in a liner relation, the Curie temperature is calculated to be 216.4 ± 0.4 K. c Zero-field-cooling (ZFC) and field-cooling (FC) curves under 100 Oe for 8-nm Fe3GeTe2 grown on sapphire substrate. For FC, with decreasing the temperature, the magnetization increases, showing an opposite trend to the ZFC curve. The Curie temperature can be roughly determined to be ~ 212.8 K, comparable to 216.4 K determined by the anomalous Hall effect measurements (Fig. 3a and b). d Magnetization hysteresis at 10 K. Due to the Al2O3 substrate, stronger diamagnetic background can be seen. After subtracting the diamagnetic background of substrate, a clear hysteresis from Fe3GeTe2 thin film is depicted in the inset with the saturation magnetization of 1.23 μB per Fe

Full size image

Superconducting quantum interference device (SQUID) was used to measure the magnetic properties. Since the easy axis of Fe3GeTe2 films is out-of-plane, the magnetic field is set to be perpendicular to the sample surface (perpendicular geometry). Figure 3c displays the Zero-field-cooling (ZFC) and field-cooling (FC) curves under 100 Oe for 8-nm Fe3GeTe2 grown on sapphire substrate. With decreasing the temperature, the magnetization under field-cooling process increases, showing an opposite trend to the ZFC curve. The Curie temperature can be roughly determined to be ~ 212.8 K, comparable to 216.4 K determined by transport Hall results. Magnetization hysteresis was achieved by scanning magnetic field back and forth in the region of ±3 Tesla at 10 K (Fig. 3d). Large diamagnetic background from Al2O3 can be clearly seen and subsequently subtracted to produce the net magnetization of Fe3GeTe2 thin films as displayed in Fig. 3d inset, with the saturation magnetization of 1.23 μB per Fe ion, smaller than that of bulk sample21, 25 ~ 1.6 μB. The shrinking behavior in the center region suggests the mixture signals from a harder magnetic phase and a slightly softer magnetic phase similar to that occurs in MnGe44 and [Co/Pd]IrMn,45 which were further proved by our Kerr microscope measurements. The detailed discussions and experimental results are provided in Supplementary information I and II.

Next, we attempt to manipulate the magnetic properties of Fe3GeTe2 thin films by systematically altering the Fe composition (Fe3+δ GeTe2) and by coupling with antiferromagnetic layer MnTe. As shown in Fig. 4, with increasing the Fe composition, the Curie temperature, extracted from the Arrott-plots (Supplementary Figs. 3–7), dramatically increases. Such a trend was also observed in Fe3+δ GeTe2 single crystals grown by self-flux reaction method,26 where the additional Fe doping decreases the lattice constant c while increases a.

Fig. 4
figure 4

Composition-dependent Curie temperature of 8 nm Fe3+δ GeTe2. These Curie temperatures of different compositions were determined by Arrott-plots of transport data. With increasing the Fe composition, the Curie temperature increases. The negligible error bar for δ = 0 is 0.4 (Fe3GeTe2, Tc = 216.4 ± 0.4 K, shown in Fig. 2). The error bars were generated from the Arrott-plots fitting process

Full size image

Besides the control of the Fe composition, the heterostructure of Fe3GeTe2/MnTe was produced to enhance the magnetic properties via the exchange interaction from the FM/AFM interface. The detailed growth conditions and characterizations of MnTe itself are presented in Supplementary Fig. 9. The structure of MnTe is hexagonal NiAs-type46 with the lattice constants a = b = 4.158 Å and c = 6.726 Å. The matching of the crystal structures enables the possibility of in situ growing MnTe directly on Fe3GeTe2. More importantly, the magnetic structure of MnTe is featured by the Mn stacking antiferromagnetically along c-axis46 that is perfectly aligned with the easy axis of Fe3GeTe2. Given the fact that the Neel temperature of MnTe is 310 K, higher than the Curie temperature of Fe3GeTe2, we have designed several periods of Fe3GeTe2/MnTe and achieved such superlattices with atomically smooth interface using MBE (see the streaky RHEED patterns in Supplementary Fig. 10). From XRD, different sets of peaks from Fe3GeTe2, MnTe and substrate Al2O3 can be observed (Fig. 5a). Figure 5b shows the R-T curve of 3-period Fe3GeTe2/MnTe with a metallic R-T behavior. Then, the angle-dependent anomalous Hall effect measurements were measured to probe the easy axis (Supplementary Fig. 11(a)) which is found to be out-of-plane. Note that hereafter the ferromagnetism measurements were always taken under the perpendicular geometry. Figure 5c displays anomalous Hall data at different temperatures. The temperature dependence of the coercive field is similar to that from the pure Fe3GeTe2. However, remarkably the coercive field is significantly enhanced (Fig. 5d). The inset of Fig. 5d are the absolute difference and increment ratios which are all above 50 % at different temperatures. At 10 K, the difference of the coercive field is as high as 50 % from 0.65 to 0.94 Tesla.

Fig. 5
figure 5

Enhanced coercive field for a heterostructure sample (Fe3GeTe2/MnTe)3. a XRD results. Peaks form Fe3GeTe2, MnTe and Al2O3 are marked in black, red, and blue, respectively. The distinguishable peaks of different class suggest the successful growth. b R-T curve of three-period Fe3GeTe2/MnTe heterostructure. The inset is a stacking illustration. c Temperature-dependent anomalous Hall results with the offset of 5 Ω. At 2.5 K, the coercive field is up to 1.1 Tesla. d Temperature-dependent coercive field of the pure Fe3GeTe2 and (Fe3GeTe2/MnTe)3. With the introduction of MnTe, the coercive field increases. The top panel in the inset shows the increment ratio and the bottom one represents the absolute enhancement

Full size image

Conventionally, when FM and AFM are coupled, the interface interaction can induce the exchange bias and the enhanced coercive field. For the exchange bias,47, 48 unlike the pure Fe3GeTe2 there will exist a shift of anomalous Hall data along x axis. To detect whether the exchange bias exists or not, we measured anomalous Hall data in the range of ±4 Tesla before which the sample is initially cooled down under ±5 Tesla from 350 K, well above the Neel temperature of MnTe. As shown in Figure S11(b)–S11(d), there is no signature of exchange bias at 10, 130, and 180 K, analogous to that occurred in Cr-(Bi,Sb)2Te3/CrSb superlattices,49 where the antiferromagnetic CrSb is magnetically ordered with spin texture altering from the bulk due to the interlay exchange coupling but without the interface exchange bias. Comparing these two systems, namely Fe3GeTe2/MnTe and Cr-(Bi,Sb)2Te3/CrSb, it is possible that the absence of the exchange bias in Fe3GeTe2/MnTe is associated with the magnetic order of MnTe (similar to CrSb), instead of antiferromagnetic state. However, such a detailed study requires neutron spectroscopy and polarized neutron reflectometry measurements that are beyond the current scope of the work.

In summary, with MBE we successfully grew wafer-scale 8 nm 2D ferromagnetic Fe3GeTe2 films. Anomalous Hall effect and butterfly-like magnetoresistance verify that the easy axis is out-of-plane. The Curie temperature of the thin film is 216.4 ± 0.4 K. Through Fe doping, the Curie temperature can be adjusted. Further combined with antiferromagnetic MnTe, the coercive field increases 50% at 10 K. With distinct magnetic properties, wafer-scale and tunable high-Curie temperature 2D FMs are promising for spintronic devices and for the study of the proximity effect when coupling with other 2D materials.

Methods

Thin film synthesis

Fe3GeTe2 thin films were grown on (0001) sapphire and (111) GaAs in a Perkin Elmer 430 MBE system with the base vacuum of 2.5 × 10−9 Torr. The substrates were cleaned using a general process before loaded into the chamber. The substrates were annealed at 600 °C for 30 min, and then decreased to the growth temperature of 340 °C. High-purity Fe (99.99%), Ge (99.999%), and Te (99.999%) at the temperature of 1165 °C, 1020 °C, and 285 °C, respectively, were co-evaporated from standard Knudsen cells. MnTe films were grown at the temperature of 320 °C with the Mn (99.99%) and Te (99.999%) cell temperature of 710 and 280 °C, respectively. The flux of each material is calibrated by the crystal oscillator. The MBE system is equipped with an in situ RHEED system.

Thin film characterizations

Structural characterizations of Fe3GeTe2 and MnTe thin films were carried out by X-ray diffraction (Bruker D8 Discover, Bruker Inc., Billerica, MA, USA) and TEM (FEI Tecnai F20) equipped with EDS. Sample composition and concentration were determined by EDS. Cross-section TEM sample was prepared by FIB (FEI Scios DualBeam).

Electrical and magnetization characterizations

The magneto-transport measurements were performed in physical properties measurement system (PPMS) by Quantum Design with a 9 T magnetic field. The devices were confined to the six-Hall-bar geometry. The data were collected using lock-in amplifiers (Stanford Research 830, Stanford Research Systems, Sunnyvale, CA, USA) in low frequency (<20 Hz). The magnetization measurements were accomplished in DC-superconducting quantum interface devices (SQUID) by quantum design with magnetic field up to 9 T.

Data availability

The authors declare that all the relevant data are available within the paper and its Supplementary Information file or from the corresponding author upon reasonable request.