The research on two-dimensional (2D) van der Waals ferroelectrics has grown substantially in the last decade. These layered materials differ from conventional thin-film oxide ferroelectrics in that the surface and interface are free from dangling bonds. Some may also possess uncommon properties, such as bandgap tunability, mechanical flexibility, and high carrier mobility, which are desirable for applications in nanoelectronics and optoelectronics. This Tutorial starts by reviewing the theoretical tools in 2D ferroelectric studies, followed by discussing the material synthesis and sample characterization. Several prototypical electronic devices with innovative functionalities will be highlighted. Readers can use this article to obtain a basic understanding of the current status, challenges, and future prospects of 2D ferroelectric materials.
Ferroelectrics are the electric counterparts of ferromagnets. At the unit cell level, ferroelectricity is the result of a spatial displacement between the center of positive charge and the center of negative charge, leading to a spontaneous polarization Ps that can be reversibly switched by an external electric field. As the ferroelectric order parameter, such an electric dipole moment is closely tied to all applications of this class of materials.1 For instance, when combined with electronic devices like field-effect transistors (FETs), the two polarized configurations are ideal for realizing nonvolatile on/off states in memory and logic applications.2 Because of electric polarization, the dielectric response of ferroelectrics can be highly nonlinear, which has found applications in tunable microwave electronics.3 The built-in electric field due to polarization charges can separate mobile electrons and holes, which is desirable for photovoltaic applications.4 As polarization is sensitively dependent on mechanical stress and the temperature, all ferroelectrics exhibit piezoelectricity and pyroelectricity. The piezoelectric effect is broadly exploited in acoustic-wave devices and actuators, whereas the coupling to thermal stimuli is utilized for sensors and power generation. In short, applications of ferroelectricity are already pervasive in modern science and technology.1–5
Compared with centuries of investigation in magnetism, the history of research in ferroelectricity is much shorter. While the first discovery of bulk ferroelectrics in Rochelle salt crystals dated back to the 1920s,6 the field mostly stayed dormant for the next three decades. The notable progress took place after the WWII when the development of the phenomenological Landau theory7 and the synthesis of barium titanate (BaTiO3) thin films8 gave a strong impetus to the study of ferroelectricity. By the 1990s, substantial improvements in the material quality and mature processing technology had enabled many practical applications of bulk and thin-film ferroelectrics.1–5 To date, the most technologically important ferroelectrics are perovskite oxides with three-dimensional (3D) lattices and strong covalent/ionic bonds. Decades of research efforts have led to remarkable functionality even when the film thickness shrinks to a few nanometers, which is highly desirable for device miniaturization.9 Nevertheless, the 3D nature of the oxide structures has intrinsic limitations. For the epitaxial growth of high-quality films, a careful selection of substrates is needed to minimize the lattice mismatch, restricting the choice of available material systems. The integration of a ferroelectric element into well-established material platforms is thus a nontrivial task. Moreover, the abrupt termination of 3D crystals at the surface would inevitably leave dangling bonds or lattice reconstruction, giving rise to a dead layer with properties different from the bulk.10 For out-of-plane polarized oxide films, another challenge is the depolarization field due to the finite screening length of surface charges, which becomes critically important as the film thickness is reduced.11 As a result, the exploration of a class of weakly bonded ferroelectric compounds beyond traditional perovskite structures is highly rewarding for both scientific and engineering communities.
Following the successful isolation of monolayer graphene from bulk graphite,12 the worldwide effort on two-dimensional (2D) layered materials provides an excellent opportunity to study intrinsic ferroelectricity in the 2D limit. These materials can be naturally cleaved at the van der Waals (vdW) gap, leaving no dangling bonds at the surface. The isolated layers can be easily placed on various substrates or stacked on each other to form heterostructures without requiring lattice match.13 In 2014, it was theoretically predicted that the distorted 1 T phase of monolayer MoS2 exhibits ferroelectric behaviors at the metal–semiconductor transition.14 In 2016, both in-plane ferroelectricity in SnTe15 and out-of-plane ferroelectricity in CuInP2S616 were experimentally verified down to one or a few atomic layers in thickness. The number of confirmed 2D ferroelectrics has been on the steady rise ever since. Subsequent investigations also reveal certain novel phenomena that are not commonly observed in thin-film oxides. To name a few, the Curie temperature Tc of few-layer SnTe is shown to be much higher than its bulk counterpart;15,17 CuInP2S6 is a rare uniaxial ferroelectric with quadruple rather than double potential well;18 α-In2Se3 exhibits unusual interlocked out-of-plane and in-plane polarization;19–24 the 1T′ phase of WTe2 shows the coexistence of spontaneous polarization and metallicity.25,26 Combined with the capability of forming vdW heterostructures, these materials can also be incorporated into prototypical devices with extraordinary performance. In short, despite the relatively short history, the study of 2D ferroelectrics has now evolved into a vibrant research field in full swing.
The four pillars of contemporary material research are theoretical calculation, material preparation, sample characterization, and device implementation. The rapid progress on 2D ferroelectrics in the past few years has benefited from the concerted efforts among these directions, as summarized schematically in Fig. 1.27–30 The purpose of this Tutorial is to provide an overview of this upcoming field such that readers may get a glimpse of the landscape of this Special Topic Collection. The rest of the article is organized as follows. Section II reviews the theoretical tools in 2D ferroelectric studies. Sections III and IV discuss the material synthesis and characterization, respectively. Section V highlights the important prototypical devices with novel functionalities, followed by conclusion and forward-looking remarks in Sec. VI. This Tutorial is by no means a comprehensive review of such a rapidly evolving area, for which the readers are referred to recent review articles.31–39
II. THEORETICAL ANALYSIS
Theoretical analysis of 2D ferroelectrics aims to determine crystal structures, quantify material parameters, and identify underlying mechanisms.32,36 The investigations are usually based on a combination of microscopic first-principles calculations and the macroscopic Landau–Ginzburg–Devonshire (LGD) theory.7 The ab initio method using density functional theory (DFT) calculates the free-energy landscape as a function of certain structural parameters, from which stable configurations of the ferroelectric and corresponding paraelectric phases are determined.40 The zero-temperature electronic polarization can be extracted from the quantum-mechanical Berry-phase calculation.40 The first-principles method can also calculate the electronic band structure and phonon dispersion spectrum. Importantly, the soft-mode theory states that the frequency of an optical phonon at the zone center becomes imaginary in the polar phase.41 If such a soft mode is indeed observed in the simulated phonon spectrum, the displacive instability is most likely the underlying mechanism of ferroelectricity. Finally, DFT calculations can provide insights into the interatomic force constant, pathway of polarization reversal, and band alignment of vdW heterostructures. The results are important for predicting new classes of 2D ferroelectrics and proposing novel device configurations.
The first-principles calculations discussed above are highly effective in determining the microscopic origin of the ferroelectric polarization at zero temperature. On the other hand, the heavy computational load makes it formidable to study phonon modes at finite temperatures and higher-order anharmonicity. Alternatively, the phenomenological LGD theory based on symmetry considerations is useful to describe the physics near a second-order phase transition.7 Here, the potential energy near the transition is expressed as the Taylor expansion of the order parameter. An external electric field can be explicitly included to reveal its influence on ferroelectricity. The parameters in the LGD theory are usually determined by fitting the results of first-principles calculations to an effective model Hamiltonian.42 Once the temperature dependence of polarization is obtained, one can directly evaluate the Curie temperature Tc, extract the critical exponent δ, and understand the universal critical phenomenon of the phase transition. The LGD theory can also be extended to describe improper ferroelectrics, the formation of domain walls, and the influences of piezoelectric and flexoelectric effects.
As a concrete example, Fei et al. performed a theoretical analysis to evaluate ferroelectricity and phase transitions of monolayer SnSe.27 The crystal structures exhibiting spontaneous polarizations (B and B′) and the phase with centrosymmetry (A) are identified by first-principles Berry-phase calculations [Fig. 2(a)]. The free-energy contour is displayed in Fig. 2(b), confirming that the two equivalent polar phases are connected by the high-symmetry nonpolar phase. The calculated phonon spectrum in Fig. 2(c) shows that the frequency of the soft mode is imaginary for the symmetry-broken phase, indicative of the displacive nature of the paraelectric–ferroelectric transition. In Fig. 2(d), spontaneous polarization Ps and energy barrier EG can be extracted from the familiar double-well potential. Finally, by fitting the results of first-principles calculations to an effective Hamiltonian and employing the Monte Carlo simulation, the temperature dependence of polarization (hence, Tc and δ) can be quantitatively evaluated under the framework of the LGD theory, as plotted in Fig. 2(e).27 Similar procedures have been applied in many theoretical works to verify the properties of known 2D ferroelectrics and predict new materials for future research.43–53
III. MATERIAL PREPARATION
To date, the list of experimentally verified vdW ferroelectrics includes Group IV monochalcogenides (SnS,54,55 SnSe,56 and SnTe),15 Group III chalcogenides (α-In2Se3,19–24 β-InSe,57 and Ga2Se358), distorted transition-metal dichalcogenides (1T′-MoTe229 and WTe2),25,26 oxy-chalcogenides (Bi2O2Se),59 transition-metal thiophosphate CuInP2S6,16,60 and layered perovskite BA2PbCl4.61 Similar to other atomically thin vdW materials, 2D ferroelectrics are usually prepared by either top-down exfoliation from bulk crystals or the bottom-up deposition of thin films. Provided that the starting material is of high quality, the famous Scotch-tape method,12 which peels off layered crystals by repeatedly applying an adhesive tape, produces flakes with the best crystallinity for fundamental research. Figures 3(a) and 3(b) display the optical image of an exfoliated MoTe2 flake29 and the AFM image of an exfoliated CuInP2S6 flake,16 respectively. Mechanical cleavage is the most effective approach when direct growth of thin films is difficult. The obvious disadvantage of mechanical exfoliation is the low yields and incompatibility with large-scale applications. Chemical exfoliation by sonication in solvents or intercalation with Li ions is also employed to isolate few-layer 2D materials with a much larger throughput.62 Because of the inferior crystalline quality and poor surface condition of the products, however, chemical exfoliation has not become a prominent method for preparing 2D ferroelectrics.
Apart from the exfoliation methods, direct growth by thin-film deposition has also been widely employed to synthesize 2D ferroelectrics. In the physical vapor deposition (PVD, also known as physical vapor transport or PVT) process, the precursor is the same as the final product.63,64 Figure 3(c) shows the schematic of PVD growth of few-layer In2Se3, where the precursor powder is placed in a quartz tube and high-purity Ar is used as the carrier gas.64 The In2Se3 vapor is transported downstream and recrystallized on the SiO2/Si substrate. The geometry and thickness of In2Se3 atomic layers can be controlled by the growth temperature and time. As seen in Fig. 3(d), triangular In2Se3 monolayers are produced at 850 °C when the growing time is 5 min. By increasing the time to 10 min, the size increases to ∼10 μm and some monolayers start to merge. After growing for 15 min, multi-layer flakes can be found. The high quality of these PVD-grown samples has been confirmed by multiple characterization tools.63,64
Different from PVD, the chemical vapor deposition (CVD) technique produces thin films through chemical reactions of two volatile precursors.65 Figure 3(e) depicts the CVD-growth of In2Se3 using Se and In2O3 powders as the precursors and H2/Ar mixture as the carrier gas.23 By carefully tuning the temperature, pressure, flow rate, and growth time, few-layer In2Se3 can be obtained on the mica substrates, as seen from the AFM image in Fig. 3(f). CVD-grown 2D ferroelectrics exhibit good crystal quality and large lateral size. In fact, wafer-scale growth of 2D materials can be achieved by the CVD technique, making it promising for applications in the electronic industry.23 Another exciting aspect of CVD is the possibility of the multi-step growth of 2D heterostructures with pristine interfaces, which is highly desirable for scientific research and device applications.
The last thin-film growth technique that has gained some momentum in 2D ferroelectric studies is molecular beam epitaxy (MBE).66 MBE is a specialized technique requiring ultrahigh-vacuum chambers and stringent surface cleaning procedures. As illustrated in Fig. 3(g), In2Se3 can be epitaxially grown on graphene using powdered In2Se3 and Se precursors.67 In contrast to the case involving strong directional covalent bonds, the lattice-matching condition is largely relaxed in the vdW epitaxy. The optimized growth temperature of 250 °C is significantly lower than that of PVD and CVD processes. The growth condition can be tuned to deposit various compositions of the In–Se compound or different phases of In2Se3. Figure 3(h) shows the AFM image of MBE-grown In2Se3 monolayer islands. As the growth time increases, the isolated flakes coalescence and form a continuous monolayer.67 Given the track record of MBE in growing high-quality thin films,66 the technique may soon produce 2D ferroelectrics with excellent material properties.
IV. SAMPLE CHARACTERIZATION
For bulk ferroelectrics, conventional characterization tools such as the Sawyer-Tower circuit can directly measure the electric polarization from the polarization–electric-field (P–E) hysteresis loop.68 For few-layer vdW ferroelectrics, however, macroscopic methods suffer from the difficulty of making parallel-plate capacitors, measuring small polarization charges, and having strong leakage current through ultrathin films. Alternatively, technologies appropriate for nanoscale characterization usually involve measurements of other properties closely related to ferroelectricity, which are elaborated below.
A. Piezo-response force microscopy (PFM)
All ferroelectrics are piezoelectrics by nature. Consequently, when it is difficult to directly access the electric polarization, probing the piezoelectric or inverse piezoelectric effect would provide a good (although indirect) alternative. In the piezo-response force microscopy (PFM) experiment,69–71 an AC electric field is applied to the piezoelectric sample through a conductive AFM tip in the contact mode. The out-of-plane expansion/retraction and in-plane shear deformation of the sample surface leads to modulated vertical and lateral motions of the cantilever. After demodulation, the PFM amplitude and phase signals reflect the magnitude of local piezoelectric response and the direction of the ferroelectric polarization, respectively. The high sensitivity of PFM is benefited from state-of-the-art force detection and lock-in measurement. The excellent spatial resolution of ∼10 nm is determined by the radius of curvature at the tip apex. Because of these attractive features, PFM has become a major tool for imaging ferroelectric domain structures in the past few decades.69–71
The PFM experiment can be operated in two modes—imaging and switching. In the imaging mode, the magnitude and phase of out-of-plane and in-plane signals are plotted as PFM images. Specifically, a 180° phase contrast would indicate the presence of ferroelectric domains. It should be noted that, while the vertical PFM reflects the out-of-plane polarization, the lateral PFM detects the torsional motion of the cantilever, which depends on the angle Φ between the cantilever axis and in-plane polarization vector. The angle-resolved PFM, as depicted in Fig. 4(a), is therefore necessary to determine the direction of in-plane polarization.61 Figure 4(b) shows the angular dependence of in-plane PFM images by rotating the cantilever long axis with respect to 2D ferroelectric BA2PbCl4 flakes.61 The polar plot shows that the PFM signals are indeed proportional to sin Φ. In the switching spectroscopy mode, the tip is positioned on a single point to perform bias-dependent PFM measurement. Figure 4(c) shows the butterfly-like amplitude loop and 180° phase reversal loop on a 4-nm-think CuInP2S6 flake.16 The ferroelectric switching is clearly observed at DC biases of +1 and −3 V. Since polarization reversal can be obtained when the applied electric field is greater than the coercive field, the PFM can also be used to write complex domain structures. In this case, it is a common practice to perform the “box-in-box” experiment [Fig. 4(d)] in order to verify the switchable behavior.23 In particular, the fact that both vertical and lateral PFM images show the same domain patterns provides strong evidence that α-In2Se3 possesses intercorrelated out-of-plane and in-plane ferroelectricity.23
Despite the widespread use of PFM in ferroelectric research, extra care is needed when performing the experiment and interpreting the data. For instance, it is important to select relatively stiff cantilevers to avoid the mixing of bending and torsional motion of the probe. Moreover, some non-ferroelectric phenomena, such as electrostatic charges on insulators or the accumulation of ions due to surface deformation, could display similar hysteresis loops or domain-like features in PFM. One can usually differentiate the short-lived electrostatic effect and the non-volatile ferroelectric effect by checking the long-term stability of the PFM results. Advanced PFM techniques are also developed to study various contrast mechanisms that lead to the piezo-response. Readers are referred to several excellent review articles for further information on this technique.69–71
B. Second-harmonic generation (SHG)
A prominent signature of ferroelectric materials is the broken inversion symmetry. Such non-centrosymmetry can be conveniently probed by the optical second-harmonic generation (SHG) technique.72 As sketched in Fig. 5(a), SHG is a nonlinear optical process in which two photons with the same frequency ω interact with the sample and generate an output photon with twice the frequency 2ω.58 Thanks to the highly sensitive optical detection, SHG offers excellent sensitivity and decent resolution (on the order of the incident light wavelength due to diffraction limit)72 for the study of nanoscale 2D flakes. The angular dependence of SHG signals can also reveal the underlying crystalline symmetry.
The polar plot in Fig. 5(b) displays the SHG polarization pattern of trilayer In2Se3 on the mica substrate under normal incidence.24 The threefold rotational symmetry of In2Se3 manifests itself through the sixfold SHG intensity pattern when the polarization of the optical excitation and detection are rotated collinearly. Further structural information about the material can be obtained by measuring the SHG pattern under slanted incidence. Moreover, the non-destructive nature of SHG makes it ideal for heating/cooling measurements across the Curie temperature, above which the 2ω signal disappears.24 Figure 5(c) shows the normalized temperature dependence of SHG intensity for CuInP2S6 flakes with a thickness of 100, 50, 30, and 10 nm, respectively.16 The vanishing of the SHG signal at ∼330 K is a strong indication of the ferroelectric–paraelectric phase transition, which involves the change from non-centrosymmetric (m) to centrosymmetric (2/m) structures for CuInP2S6.16 It should be noted that the lack of inversion symmetry is a necessary but not sufficient condition for ferroelectricity. As a result, SHG is usually combined with other techniques in the study of 2D ferroelectrics.
C. Scanning tunneling microscopy (STM)
By scanning a metal tip over the sample and monitoring the tunnel current, scanning tunneling microscopy73 (STM) is one of the most successful local probes for fundamental research. The atomic resolution obtained by STM can offer direct information on the broken inversion symmetry in ferroelectric lattices and the formation of domain structure. Through scanning tunneling spectroscopy (SPS), one can spatially resolve the energy band bending due to polarization charges. By applying a voltage pulse on the STM tip, it is also possible to manipulate the polarization direction.
Chang et al. reported a beautiful STM work on MBE-grown SnTe film with in-plane ferroelectricity.15 The STM image in Fig. 6(a) shows the formation of domain structure and the presence of domain walls. By taking atomically resolved images on both sides of the domain wall and performing fast Fourier transformation (FFT), the authors demonstrated that the lattice is distorted from a perfect square to a parallelogram, giving rise to the broken symmetry. In Fig. 6(b), spatially resolved dI/dV spectra (right panels) are taken along the lines perpendicular to the edges of two adjacent domains (left panel). The tunneling spectroscopy indicates that the energy bands shift to opposite directions, consistent with the presence of polarization charges. The temperature dependence in Fig. 6(c) confirms that the lattice distortion disappears at the Curie temperature of ∼270 K.15 In a similar work on MBE-grown monolayer SnSe, voltage pulses are applied to the STM tip [Fig. 6(d)] for the manipulation of ferroelectric domains.56 Figures 6(e)–6(g) show the dI/dV images of a ferroelectric switching sequence. During the application of voltage pulses, the domain with its polarization aligned to the external field starts to grow and eventually extends to the entire flake, consistent with the in-plane ferroelectric switching. In short, STM proves to be a powerful tool to investigate 2D ferroelectrics, although the stringent requirement on surface conditions has limited its application to date.
Before concluding this section, it is worth noting that several advanced structural characterization tools, e.g., synchrotron x-ray scattering,74 ultraviolet Raman spectroscopy,75 and scanning transmission electron microscopy (STEM),76 have been utilized to probe the broken inversion symmetry in oxide ferroelectrics. With the ever improving material quality, it is expected that these tools will also find applications in 2D ferroelectricity soon.
V. DEVICE APPLICATION
Few-layer 2D ferroelectrics differ from conventional thin-film oxide ferroelectrics in that the surface and interface are free from dangling bonds. Some vdW ferroelectrics may also possess properties that are not commonly observed in the perovskite counterparts, such as bandgap tunability, mechanical flexibility, and high carrier mobility. As a result, while the two classes of materials naturally share similar applications like memories and sensors, certain device configurations unique to or better suited for the vdW structures have been proposed and implemented, which will be briefly discussed below. Note that this section is not intended to provide a thorough list of 2D ferroelectric devices. For instance, applications based on the mechanical (mostly piezoelectric) and chemical (catalytic and gas sensing) properties are not covered here, on which the readers may refer to other articles.31,36,37
A. Ferroelectric tunnel junction diode
A ferroelectric tunnel junction77 (FTJ) diode is a two-terminal device that can switch between two reversible and non-volatile states. Polarization reversal modifies the height of the tunnel barrier, resulting in a large change of tunneling electro-resistance (TER) across the junction.78 Compared with the traditional ferroelectric random access memory, the FTJ-based memory has the advantage of non-destructive reading and electric field control of the TER. For conventional perovskite or binary-oxide-based FTJs, the limited compatibility with existing substrates leads to a small modulation of barrier height (typically, below 0.1 V) and relatively low TERs (102–106).79 In contrast, 2D ferroelectrics are compatible with various conductive substrates and metal contacts, promising high-performance device applications.
As shown in Figs. 7(a) and 7(c), FTJ diodes based on vdW ferroelectrics such as CuInP2S6,16 In2Se3,67 and MoTe229 are routinely demonstrated in the literature, although the TERs in these works are mediocre. By carefully selecting the electrode materials, Wu et al. reported an extremely high TER in the Au/Cr–CuInP2S6–graphene heterojunction [Fig. 7(d)] via barrier height modulation.30 As shown in Fig. 7(e), the current–voltage characteristics of the vdW FTJ with 4 nm CuInP2S6 and monolayer graphene exhibit a TER above 107 between the on and off states. Such an impressive TER is attributed to the unique band alignment between vdW materials, as illustrated in Fig. 7(f). For the on state, the built-in polarization field in CuInP2S6 dopes graphene to n-type and decreases the barrier height, inducing a large tunneling current. On the other hand, in the off state (accessed by an “erase” voltage of 4.5 V), the ferroelectric polarization field is switched to the opposite direction, inducing holes in graphene and substantially increasing the barrier height by ∼1 eV. It, therefore, becomes more difficult for electrons to tunnel through the junction. The authors further showed that the large effective mass in the vertical direction of CuInP2S6 can increase the TER ratio through the transmission coefficient across the tunneling barrier.30 In summary, the unique properties of both graphene and the 2D ferroelectric CuInP2S6 are crucial for achieving the high TER ratio.30
In addition to vertical FTJs perpendicular to the 2D plane, lateral FTJs are also proposed on in-plane 2D ferroelectrics such as SnTe, SnSe, and SnS.15,80,81 Due to the insufficient screening in two dimensions, the in-plane polarization could induce strong band bending around an interface. By sweeping the bias voltage beyond the switching threshold, the asymmetric contact geometry with respect to the polarization direction can lead to a large TER ratio between the on and off states. For instance, Shen et al. proposed a lateral FTJ composed of a p-type semiconductor/ferroelectric/n-type semiconductor (monolayer In:SnSe/SnSe/Sb:SnSe) homostructure.81 By switching the in-plane ferroelectricity in SnSe, quantum tunneling across the semiconducting barrier changes drastically due to charge accumulation/depletion at semiconductor/ferroelectric interfaces. The authors simulated a large TER effect of 1460% between the two polarization directions in SnSe, which comes from a dual modulation of both barrier width and barrier height. To date, the lateral FTJ is mostly discussed by theoretical analysis and device simulation,80,81 with limited experimental demonstrations. Future works are needed to implement this configuration and justify the advantage over existing designs.
B. Ferroelectric FET and ferroelectric-semiconductor FET
Ferroelectric field-effect transistors (Fe-FETs) use a ferroelectric material as the gate insulator, whose polarization state controls the conductance of the semiconducting channel.82 Because of the non-destructive readout, non-volatile memory state, and simple structure, Fe-FET is a promising memory technology for high-density integrated circuits. Moreover, the use of ferroelectric material can lead to internal amplification of the gate voltage, which may overcome the thermionic limit of sub-threshold swing (SS, 60 mV/dec at room temperature). Such Fe-FETs are also known as negative capacitance FETs (NC-FETs), which have been reported in the literature.83 In a recent study, Wang et al. demonstrated the NC-FET based on dual-gated CuInP2S6/MoS2 heterostructures.84 As a control experiment, the device configuration, transfer characteristics, and sub-threshold swing (SS) on the traditional 285 nm –SiO2 back-gate are shown in Figs. 8(a), 8(b), and 8(c), respectively. The SS under this configuration is clearly greater than the desired 60 mV/dec. On the other hand, when CuInP2S6 is used as the gate dielectric, the steep Ids − Vtg curve indicates that the SS exceeds Boltzmann's thermionic limit for a broad range of drain current. The authors further demonstrated flexible NC-FETs with sub-60 mV/dec switching characteristics under the bending radius down to 3.8 mm.84
To date, a number of groups have demonstrated Fe-FETs using 2D ferroelectrics as the gate dielectric.84–88 The depolarization field and gate leakage current, however, usually result in the short retention time of Fe-FETs. The semiconducting property of certain 2D ferroelectrics provides an alternative design to utilize ferroelectricity. As shown in Fig. 9(a), Si et al. proposed the ferroelectric semiconductor field-effect transistor (FeS-FET),89 where a ferroelectric semiconductor and a high-quality amorphous insulator are used as the channel material and gate dielectric, respectively. Since mobile charges can screen the depolarization field and the gate insulator can suppress the leakage current, this approach could offer improved performance over conventional Fe-FETs in non-volatile memory applications.
Figure 9(b) shows the device structure of FeS-FET, which takes advantage of the high carrier mobility, appropriate semiconducting gap, and room-temperature ferroelectricity of In2Se3. The fact that mobile charges, polarization switching, and non-uniform field distribution all reside in the same material leads to unique device characteristics. For instance, when 90 nm SiO2 is used as the gate insulator, the field is not strong enough to penetrate through the top surface [Fig. 9(c)]. As a result, in the polarization-down state, electrons can accumulate at the bottom surface and the channel resistance is low. Similarly, in the polarization-up state, the electron density at the bottom surface is low, resulting in high channel resistance. Since a negative voltage below the coercive threshold leads to the polarization-down state and vice versa, the ID−VGS curve exhibits a clockwise hysteresis loop, as confirmed by the experimental data in Fig. 9(d). On the contrary, when 15 nm HfO2 is used as the gate insulator, the field is sufficient to penetrate to the top surface [Fig. 9(e)]. In this case, the top surface becomes conducting due to full polarization switching, resulting in high (low) channel conductance in the polarization-down (polarization-up) state. Therefore, the ID-VGS curve shows a counterclockwise hysteresis loop, as seen in the experimental data [Fig. 9(f)]. In both cases, the FeS-FET offers a wide memory window, a high on/off ratio, and a large on-current, which are superior to conventional Fe-FETs.89,90
C. Multidirectional switchable devices
In α-In2Se3, the simultaneous in-plane and out-of-plane polarizations share the same origin in that the displacement of the center Se atom relative to its neighboring In atoms is neither perpendicular nor parallel to the 2D surface. As a result, the electrical switching of one polarization necessarily renders the reversal of the other one, which is known as the dipole locking effect.23,24 It is believed that the stabilization of ferroelectricity in In2Se3 against the depolarization field is primarily achieved by this bond configuration instead of the long-range Coulomb interaction in conventional ferroelectrics.21 The dipole locking effect can be utilized for novel electronic devices.
Dai et al. designed and fabricated a lateral switchable rectifier based on the interrelated coupling between out-of-plane and in-plane dipoles in α-In2Se3.91 As illustrated in Fig. 10(a), the out-of-plane polarization will point downward under a sufficiently large negative back-gate voltage. The in-plane polarization vector should point leftward due to the dipole locking effect. Consequently, the source–drain current can only flow to the left. The opposite is true for a sufficiently large positive back-gate voltage, under which the device behaves as a diode pointing to the right [Fig. 10(b)]. The gate-controlled bidirectional diode behavior is clearly observed in the switching characteristics plotted in Fig. 10(c). Based on such dynamic rectifying performance, a switchable half-wave rectifier can be configured [Fig. 10(d)]. As seen in Figs. 10(e) and 10(f), with a sinusoidal input waveform, the output signals are truncated according to the polarity of the gate voltage.91 In addition to the rectifier, the same dipole locking effect has been utilized in heterostructures and memristor devices.92 Since many 2D vdW materials are recently predicted to exhibit the coupled in-plane and out-of-plane polarizations,53 it is expected that more bidirectional switchable devices will be demonstrated in the near future.
D. Optoelectronic devices
When illuminated by above-gap photons, the built-in electric field in ferroelectric materials can separate the photogenerated electron–hole pairs, an effect reminiscent of that in conventional semiconductor p–n junctions.93 For perovskite ferroelectrics, applications in photovoltaics have been limited due to the large energy gaps.4 On the other hand, semiconducting vdW ferroelectrics such as α-In2Se3, β-InSe, and SnS possess small bandgaps, high carrier mobility, and decent polarization values. To date, researchers have demonstrated photodiodes, phototransistors, and optoelectronic domain-wall memories using 2D ferroelectrics.94–100
In a recent work, Xu et al. proposed and implemented an interesting optoelectronic device [Fig. 11(a)], in which optical stimuli can alter the screening field and lead to polarization reversal.99 As seen in Fig. 11(b), the dual-gated α-In2Se3 device is first prepared by an electrical pulse that sets the transistor in either the “on” (+25 V, step 1) or “off” (−25 V, step 4) state. When the device is illuminated by mild lamp light, the drain current increases dramatically (steps 2 and 5) and stays high after the light is turned off (steps 3 and 6), regardless of the previous polarization state. As illustrated in Fig. 11(c), before illumination, the screening charges on the electrodes can stabilize the up-polarization against the depolarization field. When light is shined on the sample, the photo-generated carriers will neutralize the negative and positive screening charges [Fig. 11(d)]. As a result, the depolarization field from the polarization charges and the built-in electric field due to the work function difference between graphene and gold electrodes will reverse polarization to a more stable downward position. In other words, light can erase the polarization of α-In2Se3 to the down state and reset the threshold voltage of the dual-gate device. Together with the gate control, this device can concurrently serve as a logic switch, photodetector, electronic memory, and photonic memory, as indicated in Fig. 11(e).99 These effects are of general nature to any ferroelectric semiconductors, which may open up applications in multifunctional devices.
In summary, the research progress in the field of 2D vdW ferroelectrics is breathtaking, given that the debut was less than one decade ago. It is clear that much remains to be done to fully explore the potential and opportunities of these materials. In particular, theoretical and experimental efforts are needed to identify more 2D ferroelectrics—in-plane and/or out-of-plane polarizations, high or low Curie temperatures, large or small coercive fields, semiconducting or insulating gaps—for future applications. A critically missing element in existing materials is the lack of large spontaneous polarization comparable to that of traditional oxide-based ferroelectrics. The “materials by design” approach, which combines the best of computational physics and the best of material science to shorten the development time for new products, may accelerate the discovery of such desired properties.
The success of the semiconductor industry exemplifies the importance of high-purity crystals in determining the ultimate device performance. In the meantime, a collection of advanced characterization techniques will expedite the refinement process and offer new insights. Fortunately, the synthesis and measurements of vdW ferroelectrics could ride the waves of the 2D material research in the past two decades. It is expected that 2D ferroelectrics with lower defect density will continue to emerge in the near future.
While several conceptually innovative device configurations based on few-layer vdW ferroelectrics have been proposed and implemented, as discussed above, it is fair to say that research works on practical applications are still in the early stage. Given the ability to form a clean interface with other functional 2D and 3D materials, it is possible that 2D ferroelectrics will eventually become an indispensable building block in various advanced circuitry as the control or active elements. The spontaneous polarization, piezoelectric effect, and flexible structure are desirable characteristics for multiferroic, topological, spintronic, and optoelectronic applications. Again, a multidisciplinary effort from phenomenological/computational theory, material science, chemistry, condensed matter physics, and electrical engineering will be the key to success in this burgeoning field of study.
The author acknowledges the support from the NSF Division of Materials Research Grant (No. DMR-2118806) and the Welch Foundation Grant (No. F-1814) for the preparation of this article.
Conflict of Interests
The author declares no competing financial interest.
Keji Lai: Writing – original draft (equal); Writing – review & editing (equal).
Data sharing is not applicable to this article as no new data were created or analyzed in this study.