WorksPhysics and Astronomy
Moire-Engineered Ferroelectric Transistors for Nearly Trap-free, Low-Power and Non-Volatile 2D Electronics
Abstract
Long-range moire patterns in twisted WSe2 enable a built-in, moire-length-scale ferroelectric polarization that can be directly harnessed in electronic devices. Such a built-in ferroic landscape offers a compelling means to enable ultralow-voltage and non-volatile electronic functionality in two-dimensional materials; however, achieving stable polarization control without charge trapping has remained a persistent challenge. Here, we demonstrate a moire-engineered ferroelectric field-effect transistor (FeFET) utilizing twisted WSe2 bilayers that leverages atomically clean van der Waals interfaces to achieve efficient polarization-channel coupling and trap-suppressed, ultralow-voltage operation (subthreshold swing of 64 mV per decade). The device exhibits a stable non-volatile memory window of 0.10 V and high mobility, exceeding the performance of previously reported two-dimensional FeFET and matching that of advanced silicon-based devices. In addition, capacitance-voltage spectroscopy, corroborated by self-consistent Landau-Ginzburg-Devonshire modeling, indicates ultrafast ferroelectric switching (~0.5 microseconds). These results establish moire-engineered ferroelectricity as a practical and scalable route toward ultraclean, low-power, and non-volatile 2D electronics, bridging atomistic lattice engineering with functional device architectures for next-generation memory and logic technologies.
Introduction
The relentless demand for energy efficiency in modern electronics has pushed conventional MOSFETs toward their fundamental physical limits. Historically, performance discussions have centered on the subthreshold swing (SS) [2, 13, 52, 53], which is thermally constrained to 60 mV at room temperature by the Boltzmann factor—an intrinsic “Boltzmann tyranny” [67] that broadens the OFF–ON transition and necessitates higher operating voltages. Negative-capacitance concepts proposed in 2008 [46] suggested that integrating ferroelectrics into the gate stack could amplify surface potential and yield sub-Boltzmann switching [35], motivating extensive exploration of ferroelectric oxides such as PZT and HfO-based dielectrics integrated with silicon FETs [7, 16, 19, 22, 25, 27, 29, 43, 44]. However, these approaches face persistent challenges related to interface traps, material integration, and dimensional scaling [40]. As devices continue to scale down, it has become evident that the dominant limitation is not the thermionic SS itself, but the trap-induced electrostatic inefficiencies that elevate power consumption and undermine reliable low-voltage switching.
This realization has shifted attention toward device platforms that naturally suppress interface traps while enabling ferroelectric functionality. Two-dimensional materials provide an atomically sharp, trap-clean channel with excellent electrostatic control, making them ideal hosts for ferroelectric field-effect transistors (FeFETs) in which polarization, rather than thermionic emission, modulates channel charge. Recent discoveries of intrinsic out-of-plane ferroelectricity in moiré superlattices [21, 24, 30, 58]—formed in twisted 2D bilayers such as WSe—offer a scalable and integration-friendly ferroic component fully compatible with van der Waals assembly [21, 24, 30, 58]. These ultraclean interfaces allow direct and efficient coupling of ferroelectric polarization to the channel, minimizing hysteresis from defect states and enabling stable, ultralow-voltage non-volatile operation. Such trap-suppressed ferroelectric gating provides memory windows as small as V, highlighting a promising route toward dense, low-voltage, and near-zero-standby-power memory and logic architectures.
To contextualise the significance of moiré ferroelectricity, it is essential to distinguish it from extrinsic ferroelectric gating using oxide or doped hafnia layers. Unlike externally deposited ferroelectrics, moiré-induced polarization arises intrinsically from periodic lattice reconstruction, leading to stable dipole ordering without chemical modification or epitaxial strain [55, 57]. This intrinsic nature provides an atomically clean route to integrate negative capacitance within van der Waals heterostructures, potentially lowering while preserving device scalability. In recent years, several research groups have made significant progress in addressing the integration challenges discussed above, thereby unlocking access to a range of emergent physical phenomena. Building on these advances, we have conducted a systematic investigation of the electronic and optoelectronic responses in 2D TMDC-based ferroelectric heterostructures, employing graphene as a sensing layer [15, 47].
Here, we demonstrate a two-dimensional moiré ferroelectric field-effect transistor (FeFET) based on a twisted WSe bilayer, which harnesses intrinsic polarization to achieve robust and improved device performance. Our FeFET exhibits a near-ideal subthreshold swing of mV dec with minimal trap-induced broadening and a stable memory window of 0.10 V. Self-consistent Landau-Ginzburg-Devonshire modelling quantifies the polarization-field relation, linking the observed hysteresis to fast ferroelectric switching dynamics in twisted WSe. Through sweep-rate-dependent hysteresis measurements, we probe the kinetics of ferroelectric domain switching and reveal that it is governed by a domain-wall-limited mechanism consistent with moiré-induced pinning. Capacitance spectra show a pronounced ferroelectric switching peak in , absent in control devices, and four distinct regimes—accumulation, depletion transition, ferroelectric switching, and inversion (deep depletion)—whose boundaries shift with frequency. Furthermore, we employ an equivalent-circuit model to rigorously analyse broadband capacitance–voltage spectroscopy data, enabling us to delineate the characteristic relaxation times of the ferroelectric layer ( s) from those of distinct interface trap populations. The quantum-capacitance slope reflects the gate-modulated density of states of MoS, while the low parallel conductance quantifies dielectric losses. This work establishes a quantitative framework for understanding the interplay between intrinsic moiré-engineered ferroelectricity and extrinsic interfacial dynamics, paving the way for low-power, steep-slope logic and nonvolatile memory [36] based on designer two-dimensional materials.
Experimental Section
Atomically thin layers of MoS, WSe, graphite, and hBN were mechanically exfoliated onto SiO (285 nm)/Si substrates using the Scotch-tape method. Monolayers of MoS and WSe were identified by optical contrast and confirmed via Raman spectroscopy. To assemble the heterostructures, we employed a deterministic dry-transfer technique using a polycarbonate (PC) film on a polydimethylsiloxane (PDMS) hemispherical stamp, which enabled precise pick-up and release of individual flakes. The device stack (Figure 1a) consisted of MoS (channel) / hBN (dielectric, thickness nm) / twisted bilayer (TBL) WSe (ferroelectric) / graphite (back gate). The rotational alignment of the two WSe layers was controlled by a high-precision rotational stage, achieving twist angles of , as established in Refs. [8, 9, 10]. Electrical contacts to the MoS channel were defined using electron-beam lithography followed by Cr/Au (5 nm/50 nm) deposition, ensuring low-resistance ohmic contacts (Figure 1a). Two nominally identical FeFET devices with comparable geometry were fabricated and characterized. For comparison, control devices were also fabricated: (i) MoS/hBN/graphite FETs without the TBL WSe layer (see Supplementary Information), to isolate the ferroelectric contribution, and (ii) TBL WSe/hBN/graphite stacks for piezoresponse force microscopy (PFM) to confirm ferroelectric polarization switching. All electrical measurements were carried out under ambient conditions using a Keithley 4200 semiconductor parameter analyzer, and capacitance–voltage characteristics were measured over the 1 kHz–1 MHz frequency range with the CVU module.
Figure 1. Moiré ferroelectricity and device characteristics (a) Optical micrograph of a representative MoS/hBN/twisted-WSe/graphite device. (b) Illustration of ferroelectric domains in twisted WSe. Green and blue colors represent up and down polarization, respectively. (c) Transfer characteristics (–) show a narrow clockwise hysteresis window ( V), while the control MoS/hBN/graphite device exhibits negligible hysteresis, excluding interface traps as the dominant mechanism (c, inset). (d) A schematic of the FeFET stack and the equivalent capacitance model is shown. (e) The calculated polarization–electric field (–) loop using the Landau–Khalatnikov formalism matches the extracted remanent polarization ( µC cm) and coercive field ( V nm), while the corresponding Landau free-energy profile (f) illustrates the characteristic double-well potential, confirming robust ferroelectric switching in twisted WSe.
Ferroelectric Domain and Electrical Hysteresis
A strong piezoelectric response is observed in the twisted WSe layer, as it is revealed by piezoresponse force microscopy in our previous work [47] and highlights the existence of oppositely polarized domains in twisted TMDC (S6). A schematic in this regard is also described, the polar features as illustrated in Figure 1b. The topography collected at high magnification clearly resolves alternate triangular domains of opposite phase, separated by non-polarized regions, confirming the presence of moiré-induced ferroelectric domains. While these static domain patterns establish the microscopic origin of polarization, their functional relevance becomes evident in the device transfer characteristics (Figure 1c), where the reversible switching of these domains manifests as a well-defined but narrow hysteresis window in the – curve with a positive threshold-voltage shift between forward and reverse sweeps (). Unlike conventional FeFETs [17, 20, 62] where counterclockwise hysteresis arises from full ferroelectric switching in an n-type channel, the direction observed here reflects the unique field distribution [49] across the MoS–hBN–twisted-WSe stack. The finite thickness of the hBN dielectric screens the effective ferroelectric bias , enabling only partial polarization switching during sweeps. The moiré potential landscape remains with a broad coercive-field distribution , such that the forward sweep () reverses solely low- domains, while the reverse sweep () benefits from residual polarization and inverted screening to activate higher- domains, yielding a net threshold shift
where is the coercive-field distribution arising from moiré disorder, and reflects domain-wall mobility. Importantly, control devices without the twisted-WSe layer, but MoS on hBN alone, show negligible hysteresis, ruling out interface traps as the dominant mechanism (inset of Figure 1c). The clean MoS/hBN interface, in contrast to MoS/SiO devices where large trap-induced clockwise [1, 26] loops are typically observed, further supports a ferroelectric origin. The observed hysteresis window corresponds to polarization switching between two stable remanent states. Using the equivalent series-capacitance model of the stack (schematic in Figure 1d), the remanent polarization is extracted as , yielding , corresponding to . The coercive field is estimated as , giving . Those estimated values from measurement are in good agreement with previous reports [55, 60]. These values, together with the narrow yet reproducible hysteresis window of , combined with stable remanent polarization, demonstrate excellent non-volatile memory reliability in our moiré ferroelectric transistors.
Self-Consistent Polarization Dynamics
To quantitatively connect these nanoscale ferroelectric signatures with device operation, we analyzed the polarization–electric field (–) characteristics of the MoS/hBN/twisted-WSe stack using Landau–Ginzburg–Devonshire (LGD) theory [48]. The ferroelectric free energy density is expressed as , where the Landau coefficients are extracted by fitting to the experimental memory window and capacitance model. The field across the ferroelectric, , is obtained self-consistently by combining LGD theory with electrostatics. The displacement field is given by , while the total gate voltage divides across the dielectric and ferroelectric as . Solving these relations yields
which shows that the effective field is screened as polarization increases. At equilibrium, the LGD restoring field balances this electrostatic field, giving a nonlinear equation in that we solve numerically using the Newton–Raphson method. Here, the ferroelectric free-energy density is modeled following the Landau–Ginzburg–Devonshire (LGD) formalism [18, 34]:
where is the polarization, and , , and are material-specific coefficients. For twisted WSe,
These parameters were extracted by fitting to the experimental memory window and capacitance spectra. The use of a sixth-order expansion ensures convergence near the coercive field and prevents unphysical polarization divergence under high bias. Absorbing those extracted parameters into the equation, we simulate the – curve reflected in Figure 1e, which highlights the coercive field of and remanent polarization of . In addition, the simulated polarization–electric field (–) curve exhibits a slight deviation from the ideal S-shaped hysteresis typically observed in ferroelectric materials. This deviation reflects the presence of a relatively shallow double-well potential landscape (as illustrated in Figure 1f), which limits the accessibility of the negative-capacitance regime in our device structure due to capacitance mismatch.
Ferroelectric Control of Subthreshold Swing
Building on the hysteresis behavior, the impact of moiré-engineered ferroelectricity on the switching characteristics is revealed through subthreshold swing (SS) measurements. Figure 2a–Figure 2c present the transfer characteristics, revealing a hysteresis that remains essentially independent of , accompanied by a progressive improvement in subthreshold swing (SS) from to mV with increasing . Figure 2d shows the point-by-point subthreshold swing (SS) as a function of drain current, revealing an extended current range over which the SS approaches the fundamental 60 mV limit. Forward and backward sweeps yield nearly identical SS values, with only a marginal advantage for the backward sweep. In ferroelectric transistors where traps dominate, backward sweeps typically show a pronounced SS improvement due to trap neutralization followed by enhanced ferroelectric polarization. The absence of such a signature here indicates that trap states are minimally involved in the hysteresis mechanism. Instead, the bias-independent hysteresis suggests that the twisted-WSe ferroelectric layer maintains a stable polarization over the measurement cycle, with negligible screening by trapped charges in the gate stack heterostructure.

Figure 2. Drain-bias dependence of subthreshold characteristics in 2D FETs. (a–c) Transfer curves at increasing () showing stable hysteresis and progressive improvement in subthreshold swing (SS). (d) Point-by-point SS versus drain current (), demonstrating approach to the near-ideal 60 mV dec limit across a broad range. (e) Comparison with a MoS/hBN control device, highlighting the superior SS (64 mV dec) in the FeFET versus 130 mV dec in the control.
Further confirmation arises from our previous work Ref. [47] using dual-gate measurements on a graphene channel, where the top-gate dielectric (hBN) exhibits no detectable hysteresis, while the back-gate incorporating twisted WSe shows a clear hysteresis loop. This contrast unambiguously attributes the hysteresis to the ferroelectric polarization of the twisted WSe, rather than trap-assisted effects. Compared to a reference MoS/hBN device without the twisted-WSe interlayer, which shows SS values exceeding 130 mV , the present device approaches the thermionic limit of 60 mV across much of the subthreshold regime. The SS variation with drain current remains close to ideal, indicating strong gate control and minimal trap-induced broadening of the subthreshold slope. This highlights the role of the twisted-WSe layer in enabling stable, low-trap, nonvolatile electrostatic modulation of the MoS channel. To assess device uniformity, Fig. S3 and S4 (Supplementary Information) presents transfer characteristics for three FeFETs fabricated under identical conditions, exhibiting subthreshold swings of 63–66 mV and memory windows between 0.09–0.11 V. This consistency confirms the reproducibility of the fabrication method.
Rate-Dependent Switching Kinetics
To further probe the origin of hysteresis, we performed sweep-rate–dependent transfer measurements of the FeFET. Figure 3a shows that the hysteresis width () systematically decreases with increasing sweep rate, consistent with a domain-wall–limited ferroelectric switching process rather than trap-controlled dynamics. The dependence of on sweep rate () is well captured by a stretched-exponential formalism [23, 63], , where is the saturation hysteresis at vanishing sweep rate, the critical rate at which the switching fraction is reduced by , and characterizes the breadth of the domain-wall velocity distribution. Experimental data (Figure 3b, symbols) are well reproduced by the fitted curve, yielding V, V h, and . The large indicates nearly complete polarization reversal under quasi-static biasing, whereas the small suggests moderate domain-wall mobility with switching suppressed even at relatively slow ramp rates. The broad exponent () implies a wide distribution of domain-wall velocities, consistent with pinning from moiré-induced strain fields or interfacial disorder. Together, these results confirm that the observed hysteresis originates predominantly from intrinsic ferroelectric switching, with charge trapping playing only a negligible role.

Figure 3. Sweep-rate dependence of hysteresis dynamics. (a) Transfer curves (–) measured at different sweep rates, shifted by 1 V for clarity, showing systematic reduction of hysteresis with increasing rate. (b) Extracted hysteresis width versus sweep rate, fitted using a domain-wall-limited switching model.
Capacitance Spectroscopy of 2D FET
To elucidate the underlying charge modulation mechanisms, capacitance–voltage (–) measurements were performed on MoS/hBN devices both with and without the ferroelectric twisted WSe layer in the gate stack. The FeFET configuration (MoS/hBN/twisted-WSe) exhibited a pronounced enhancement in capacitance (Figure 4a) compared to the control MOSFET (MoS/hBN). This enhancement is consistent with the additional charge contribution from polarization switching in the ferroelectric layer. The derivative serves as a sensitive probe for this switching [18, 23], revealing a single, sharp, and symmetric peak in the FeFET curve (Figure 4b) with a full width at half maximum (FWHM) of 0.80 V and a peak amplitude of 2.3 pF/V. This signature is a direct fingerprint of domain-wall-mediated ferroelectric switching. In contrast, the control MOSFET displayed a much broader and less intense peak (FWHM = 1.14 V, amplitude = 0.95 pF/V), characteristic of non-cooperative, trap-mediated electrostatic modulation. This is consistent with a conventional MOS capacitor structure, where the capacitance transitions from accumulation to depletion and eventually saturates at high negative , representing the deep depletion or inversion capacitance.
Figure 4. CV spectroscopy of 2D FeFETs. (a) CV characteristics of the FeFET compared with a control device, showing enhanced capacitance (inset: device schematic). (b) Differential plot displaying a sharp peak for the twisted-WSe device, indicative of ferroelectric switching, whereas the control shows a broad peak corresponding to deep depletion or trap response; Lorentzian fits are used to extract the FWHM. (c) Frequency-dependent CV curves of another FeFET reveal an additional ferroelectric switching regime beyond the conventional CV response. (d) as a function of frequency showing kinetic limitation in domain wall motion. (e) Experimental as a function of frequency at different values. (f) Extracted and as a function of . (g) Experimental and simulated two-dimensional CV spectroscopy maps, highlighting the agreement between measurement and self-consistent modeling.
Further insights are gained from the device’s frequency-dependent response. At low frequencies (30 kHz), the – curve exhibits four characteristic regimes (Figure 4c): (i) a high-capacitance accumulation region (3 to 5 V), (ii) a depletion transition (1 to 3 V), (iii) a ferroelectric switching window ( V to 1 V) where polarization reversal in the twisted WSe layer induces negative capacitance effects, followed by (iv) a post-switching decrease in capacitance from V to V, signifying completion of ferroelectric switching as the depletion region expands into the semiconductor channel. At high negative gate voltages ( V), the device is either in strong inversion (formation of a hole inversion layer in the MoS/WSe channel) or deep depletion. The total capacitance of the FeFET significantly decreases from a maximum of 5.25 pF at 1 kHz to 1.8 pF at 1 MHz. This frequency suppression is a hallmark of relatively slow-response phenomena, namely ferroelectric polarization reversal and the contribution of interface trap states. Simultaneously, the ferroelectric switching peak in vanishes at high frequencies, indicating kinetic limitations in domain-wall motion under rapid gate modulation (Figure 4d). Above several hundred kilohertz, the device response converges to the dielectric-only limit, confirming the ferroelectric origin of the low-frequency enhancement. From the minimum capacitance in the subthreshold regime, we infer that the capacitance floor is set by the depletion capacitance , which remains finite even at low carrier densities. The absence of a vanishing capacitance rules out a dominant quantum capacitance contribution in this regime, as such an effect would appear in series with and further reduce the total capacitance. To gain a deeper understanding of the total measured capacitance, , in a gated two-dimensional semiconductor device is modeled as a series combination of the oxide capacitance of the top-gate , the quantum capacitance , and the frequency-dependent interface trap capacitance :
The interface-trap contribution is represented by two characteristic branches with distinct time constants and , expressed as
where and are the trap densities (in ), is the elementary charge, and is the angular frequency. The extracted quantum capacitance and the characteristic trap time obtained from the constrained global fit are plotted as a function of in Figure 4(f).
Trap Density analysis
The interface trap density () near the semiconductor–dielectric interface was first estimated from the subthreshold swing (SS) [20] characteristics using
which reflects the trap density at the Fermi level (Figure 4(g)) when the device is near turn-on. For the twisted–WSe FeFET, the extracted mV/dec corresponds to cm eV, whereas the control device without the ferroelectric layer ( mV/dec) yields cm eV. This one-order-of-magnitude reduction in interface trap density indicates an ultraclean interface in the twisted–WSe gated structure, consistent with the suppression of interfacial disorder and charge scattering. To further validate the trap response, high–low frequency [61, 66] – spectroscopy was performed. Using the standard formulation [61]
the extracted values (Figure 4e) were slightly higher than those obtained from the SS analysis, consistent with additional slow interface states that respond at lower frequencies. Frequency-dependent capacitance measurements at various gate voltages revealed a linear dispersion with beyond the ferroelectric switching regime, indicating that the apparent frequency response in the accumulation region cannot be attributed to border traps. Instead, the observed dispersion originates from ferroelectric polarization dynamics. This ferroelectric contribution was quantitatively modeled using a constrained Cole–Cole magnitude formalism to extract the characteristic relaxation time () and dispersion parameter () (see Supplementary Information).
Equivalent Circuit Analysis
The equivalent circuit employed to analyze the small-signal response of the top-gated monolayer MoS field-effect transistors (FETs) is illustrated in fig:5a along with the device schematic in fig:5b. The model comprehensively captures both intrinsic and extrinsic contributions to the total measured capacitance, enabling a quantitative interpretation of parasitic effects and channel dynamics. The circuit consists of the intrinsic gate-stack capacitances—namely, the oxide capacitance () in series with the quantum capacitance ()—in parallel with the interface trap branch characterized by the capacitance () and resistance (), which represent carrier capture and emission processes at the dielectric/semiconductor interface. The total device resistance is separated into the gate-modulated channel resistance () under the gate and the bias-independent access resistance (), which includes both the MoS sheet resistance in the ungated region and the metal–semiconductor contact resistance.
To account for the finite that limits the measured capacitance at high frequencies, the experimental response was modelled using an equivalent parallel configuration [11] given by
where represents the ideal gate capacitance in the absence of interface traps and parasitic effects, and is the angular frequency. This formulation captures the attenuation of the apparent capacitance in the accumulation regime as increases. The simulation neglects other resistive components such as , which are effectively shunted under strong accumulation. The extracted (Figure 5c) asymptotically approaches at 1 MHz when , validating the consistency of the equivalent circuit model and confirming that the measured suppression of the capacitance in the accumulation regime primarily originates from the series resistance in the ungated MoS access regions.

Figure 5. Simplified equivalent circuits to model MoS–FeFET –. (a) Full equivalent circuit illustrating the formation of the total measured capacitance . (b) Schematic of the capacitance measurement configuration, where both the source and drain electrodes are connected to the low terminal, and the top gate is connected to the high terminal. (c) Experimental and simulated total capacitance–voltage (–) characteristics of the MoS FeFET, showing good agreement in the accumulation regime. (d) Experimental and simulated two-dimensional – spectroscopy maps, highlighting the consistency between measurement and self-consistent modeling.
Quantum capacitance analysis
Based on the ideal equivalent circuit model [3] in the high-frequency limit, where the interface trap capacitance () is negligible, the theoretical correlation between the channel potential () and the applied gate voltage () can be expressed as
where is a fitting parameter corresponding to the gate voltage at which the Fermi level lies at the mid-gap ( eV). This term effectively accounts for the intrinsic -type doping present in monolayer MoS. By combining this relation with the theoretically derived expression of the quantum capacitance (), the variation of as a function of the gate voltage () (Figure 5c inset) can be quantitatively determined and then compared with a microscopic two-dimensional Fermi-occupation model. Experimentally, the oxide capacitance was estimated from the accumulation plateau, and the quantum capacitance was recovered using the series relation
The theoretical curve was computed for a 2-D conduction band with a constant band-edge density of states
using , and a conduction-band edge eV referenced to the midgap . The sheet density and quantum capacitance were evaluated from Fermi statistics as
This combined procedure provides a direct, physically transparent test of the DOS-derived quantum capacitance and provides information on interface traps, incomplete accumulation, or mis-estimated oxide coupling — immediately apparent, offering useful diagnostics for 2-D FeFET/TFET device modelling.
To quantitatively interpret the capacitance–voltage characteristics, we modeled the system using an equivalent circuit [18, 34] that accounts for ferroelectric switching (Figure 4d), interface traps, quantum capacitance, and conductance losses:
The equivalent-circuit model [12] (Figure 5d) accounts for ferroelectric polarization (), interface traps (, ), and quantum capacitance (), enabling a self-consistent interpretation of the – response [18]. The extracted relaxation times ( µs, s, s) separate intrinsic ferroelectric switching from interface-mediated processes, emphasizing that the dominant kinetics originate from domain-wall motion within the moiré potential landscape. The fitting uncertainty of the extracted relaxation times was within ±10% across multiple frequency sweeps, and repeated CV measurements yielded consistent spectra, confirming that the observed frequency dependence originates from intrinsic ferroelectric dynamics rather than measurement artifacts.
Benchmarking Device performance
The overall device performance of the twisted–WSe FeFET was benchmarked in terms of both switching steepness and polarization strength (Figure 5). The device exhibits a subthreshold swing of 64 mV dec, approaching the sub-Boltzmann limit, along with a remanent polarization of µC/cm and a coercive field of V/nm. These values substantially outperform the control MoS/hBN transistor ( mV dec) and compare favorably with state-of-the-art twisted or moiré ferroelectric transistors based on InSe, hBN, and MoS heterostructures. The enhanced charge–polarization coupling and improved electrostatic control underscore the effectiveness of moiré engineering in achieving steep-slope and low-voltage operation in two-dimensional ferroelectric field-effect transistors.
A comparative analysis of reported moiré ferroelectric systems reveals that the polarization magnitude and coercive field exhibit strong dependence on interlayer coupling and twist angle. The relatively high remanent polarization observed in the twisted–WSe device arises from enhanced interfacial dipole alignment and efficient domain-wall motion across the moiré potential landscape. Simultaneously, the moderate coercive field facilitates low-power switching without polarization fatigue. Together, these characteristics place the present device among the best-performing van der Waals ferroelectric transistors reported to date, bridging the gap between steep-slope logic and nonvolatile memory functionalities within a unified two-dimensional platform.
Conclusion
Our findings reveal that moiré engineering in 2D heterostructures can deliver simultaneously steep-slope switching and robust non-volatile memory, directly addressing the energy-delay trade-off in future electronics. The switching dynamics—driven by domain-wall motion pinned within the moiré potential—combine rapid response ( relaxation) with long-term stability, a balance essential for both high-speed logic and reliable data retention. By disentangling ferroelectric and interfacial effects through capacitance–voltage spectroscopy, we establish a quantitative framework for rational FeFET design. Our findings open a route toward integrating moiré ferroelectrics with scalable CMOS-compatible architectures. Future work should explore replacing hBN with high- dielectrics to enhance gate coupling, as well as exploring heterobilayer combinations (e.g., ) for tunable polarization landscapes. Such advancements could enable sub-50 mV dec switching and multi-bit memory functionalities with reconfigurable neuromorphic circuits, positioning moiré-engineered FeFETs as key candidates for beyond-Boltzmann logic circuits.
Experimental method
All electrical measurements were performed in an ultrahigh vacuum cryogenic probe station ( mbar) using a Keithley 4200 semiconductor parameter analyzer with high-sensitivity triaxial cable to reduce noise and ultralow current measurements of 0.1 fA resolution. Capacitance–voltage (–) measurements spanned frequencies from 1 kHz to 1 MHz using the CVU module, with the device grounded at source and back gate held at a constant potential. To ensure reproducibility, each measurement was repeated at least three times, and device statistics were compiled across all runs. Supplementary Information includes raw transfer curves, control device measurements, and details of instrument calibration.
Supplementary Information
Supplementary data include: (i) raw – and – characteristics for all FeFETs, (ii) parameter fitting workflow for the LGD and equivalent-circuit models, (iii) additional PFM confirmation of domain orientation, and (iv) details of the Self-Consistent Polarization Model and Capacitance spectroscopy.
Author Contributions
R.D. and A.S. designed all the experiments. A.S. and R.D. completed fabrication and characterization, A.S. and R.D. performed the measurements. R.D. and A.S. wrote the manuscript and analyzed the data, and performed device simulations. S.S. contributed to reviewing and editing the manuscript. A.G. supervised the overall project, edited, and revised the manuscript. All authors discussed the results and commented on the manuscript.
Code availability
The computer code used to generate the results reported in this study is available from the corresponding author upon request.
Competing interests
The authors declare no conflict of interest.
ACKNOWLEDGMENTS
The authors acknowledge the usage of NNFC facilities at CeNSE, IISc, and funding from the Department of Science and Technology (DST), Govt. of India. A.G. acknowledges the J.C. Bose Fellowship (Grant number SP/DSTO-18-2038) from Science and Engineering Research Board, DST and Nano Mission, DST, Govt. of India for financial support under Grant No. DST/NM/TUE/QM-10/2019. S.S. would like to acknowledge MEITY, Govt. of India (Project Number: SP/MEIT-24-0007) for financial support. K.W. and T.T. acknowledge support from the JSPS KAKENHI (Grant Numbers 21H05233 and 23H02052), the CREST (JPMJCR24A5), JST and World Premier International Research Center Initiative (WPI), MEXT, Japan.

Figure 6. Benchmarking subthreshold swing in Moiré-engineered ferroelectric transistors. (a) Comparison of the subthreshold swing (SS) of our twisted-WSe FeFETs with reported 2D FETs and 2D FeFETs [?, 4, 5, 14, 28, 31, 33, 37–39, 41, 42, 45, 50, 51, 54, 56, 59, 64, 65], highlighting near-ideal SS enabled by moiré-induced ferroelectric coupling and efficient polarization switching. (b) The twisted–WSe device exhibits a remanent polarization of and a coercive field of , comparable to or exceeding values reported for other twisted or moiré ferroelectric systems, confirming efficient interlayer coupling and robust polarization [6, 7, 32, 35, 49, 55, 60].
References
- [1]F. Ali, F. Ahmed, M. Taqi, S. B. Mitta, T. D. Ngo, D. J. Eom, K. Watanabe, T. Taniguchi, H. Kim, E. Hwang, et al. Traps at the hbn/wse2 interface and their impact on polarity transition in wse2. 2D Materials, 8(3):035027, 2021.
- [2]L. Chang, D. J. Frank, R. K. Montoye, S. J. Koester, B. L. Ji, P. W. Coteus, R. H. Dennard, and W. Haensch. Practical strategies for power-efficient computing technologies. Proceedings of the IEEE, 98(2):215–236, 2010.DOI
- [3]L. Chen, Z. Lu, Y. Song, R. Guo, H. Li, J.-P. Raskin, D. Flandre, Y. Liu, L. Liao, and G. Li. Quantum capacitance impact on low-frequency noise in mo s 2 transistors. Physical Review Applied, 23(5):054012, 2025.DOI
- [6]W. Chung, M. Si, P. R. Shrestha, J. P. Campbell, K. P. Cheung, and P. D. Ye. First direct experimental studies of hf 0.5 zr 0.5 o 2 ferroelectric polarization switching down to 100-picosecond in sub-60mv/dec germanium ferroelectric nanowire fets. In 2018 IEEE Symposium on VLSI Technology, pages 89–90. IEEE, 2018.
- [7]S. Dasgupta, A. Rajashekhar, K. Majumdar, N. Agrawal, A. Razavieh, S. Trolier-Mckinstry, and S. Datta. Sub-kT/q switching in strong inversion in pbzr 0.52 ti 0.48 o 3 gated negative capacitance fets. IEEE Journal on Exploratory Solid-State Computational Devices and Circuits, 1:43–48, 2015.
- [8]R. Debnath, I. Maity, R. Biswas, V. Raghunathan, M. Jain, and A. Ghosh. Evolution of high-frequency raman modes and their doping dependence in twisted bilayer mos 2. Nanoscale, 12(33):17272–17280, 2020.DOI
- [9]R. Debnath, S. Sett, R. Biswas, V. Raghunathan, and A. Ghosh. A simple fabrication strategy for orientationally accurate twisted heterostructures. Nanotechnology, 32(45):455705, 2021.DOI
- [10]R. Debnath, S. Sett, S. Kundu, R. Biswas, V. Raghunathan, M. Jain, A. Ghosh, and A. Singh. Tuning exciton complexes in twisted bilayer wse 2 at intermediate misorientation. Physical Review B, 106(12):125409, 2022.DOI
- [11]N. Fang and K. Nagashio. Accumulation-mode two-dimensional field-effect transistor: Operation mechanism and thickness scaling rule. ACS Applied Materials & Interfaces, 10(38):32355–32364, 2018.DOI
- [12]N. Fang and K. Nagashio. Band tail interface states and quantum capacitance in a mono-layer molybdenum disulfide field-effect-transistor. Journal of Physics D: Applied Physics, 51(6):065110, 2018.
- [13]A. D. Franklin. Nanomaterials in transistors: From high-performance to thin-film applications. Science, 349(6249):aab2750, 2015.DOI
- [15]N. K. Gill, S. Sett, R. Debnath, A. Singha, K. Watanabe, T. Taniguchi, and A. Ghosh. Moiré ferroelectricity-enhanced optoelectronic response in an all-2d van der waals hybrid. Small, page e05797, 2025.DOI
- [16]I. Ham, Y. Jeong, S. J. Baik, and M. Kang. Ferroelectric polarization aided low voltage operation of 3d nand flash memories. Electronics, 10(1):38, 2020.DOI
- [17]G. He, H. Ramamoorthy, C.-P. Kwan, Y.-H. Lee, J. Nathawat, R. Somphonsane, M. Matsunaga, A. Higuchi, T. Yamanaka, N. Aoki, et al. Thermally assisted nonvolatile memory in monolayer mos2 transistors. Nano letters, 16(10):6445–6451, 2016.pubmed.ncbi.nlm.nih.gov/27680095
- [18]M. Hoffmann, A. I. Khan, C. Serrao, Z. Lu, S. Salahuddin, M. Pešić, S. Slesazeck, U. Schroeder, and T. Mikolajick. Ferroelectric negative capacitance domain dynamics. Journal of Applied Physics, 123(18), 2018.DOI
- [19]A. Jain and M. A. Alam. Proposal of a hysteresis-free zero subthreshold swing field-effect transistor. IEEE Transactions on Electron Devices, 61(10):3546–3552, 2014.DOI
- [20]H. Jeon, S.-G. Kim, J. Park, S.-H. Kim, E. Park, J. Kim, and H.-Y. Yu. Hysteresis modulation on van der waals-based ferroelectric field-effect transistor by interfacial passivation technique and its application in optic neural networks. Small, 16(49):2004371, 2020.DOI
- [21]S. Jiang, R. Du, J. Jiang, G. Liu, J. Huang, Y. Du, Y. Han, J. Xiao, D. Zhang, F. Lian, et al. The interplay of ferroelectricity and magneto-transport in non-magnetic moiré superlattices. Nature Communications, 16(1):5640, 2025.DOI
- [22]J. Jo and C. Shin. Negative capacitance field effect transistor with hysteresis-free sub-60-mv/decade switching. IEEE Electron Device Letters, 37(3):245–248, 2016.DOI
- [23]J. Y. Jo, S. M. Yang, T. Kim, H. N. Lee, J.-G. Yoon, S. Park, Y. Jo, M. Jung, and T. W. Noh. Nonlinear dynamics of domain-wall propagation in epitaxial ferroelectric thin films. Physical review letters, 102(4):045701, 2009.DOI
- [24]S. Kamaei, X. Liu, A. Saeidi, Y. Wei, C. Gastaldi, J. Brugger, and A. M. Ionescu. Ferroelectric gating of two-dimensional semiconductors for the integration of steep-slope logic and neuromorphic devices. Nature Electronics, 6(9):658–668, 2023.DOI
- [25]K. Karda, A. Jain, C. Mouli, and M. A. Alam. An anti-ferroelectric gated landau transistor to achieve sub-60 mv/dec switching at low voltage and high speed. Applied Physics Letters, 106(16), 2015.DOI
- [26]N. Kaushik, D. M. Mackenzie, K. Thakar, N. Goyal, B. Mukherjee, P. Boggild, D. H. Petersen, and S. Lodha. Reversible hysteresis inversion in mos2 field effect transistors. npj 2D Materials and Applications, 1(1):34, 2017.DOI
- [27]A. I. Khan, C. W. Yeung, C. Hu, and S. Salahuddin. Ferroelectric negative capacitance mosfet: Capacitance tuning & antiferroelectric operation. In 2011 International Electron Devices Meeting, pages 11–3. IEEE, 2011.DOI
- [29]K. Li, Y. Peng, W. Xiao, F. Liu, Y. Zhang, Z. Feng, H. Dong, Y. Liu, Y. Hao, and G. Han. A comparative study on the polarization, reliability, and switching dynamics of hfo 2-zro 2-hfo 2 and zro 2-hfo 2-zro 2 superlattice ferroelectric films. IEEE Transactions on Electron Devices, 70(4):1802–1807, 2023.DOI
- [30]X. Li, B. Qin, Y. Wang, Y. Xi, Z. Huang, M. Zhao, Y. Peng, Z. Chen, Z. Pan, J. Zhu, et al. Sliding ferroelectric memories and synapses based on rhombohedral-stacked bilayer mos2. Nature Communications, 15(1):10921, 2024.DOI
- [32]X. Liu, D. Wang, K.-H. Kim, K. Katti, J. Zheng, P. Musavigharavi, J. Miao, E. A. Stach, R. H. Olsson III, and D. Jariwala. Post-cmos compatible aluminum scandium nitride/ 2d channel ferroelectric field-effect-transistor memory. Nano Letters, 21(9):3753–3761, 2021.DOI
- [34]K. Majumdar, S. Datta, and S. P. Rao. Revisiting the theory of ferroelectric negative capacitance. IEEE Transactions on Electron Devices, 63(5):2043–2049, 2016.
- [35]F. A. McGuire, Y.-C. Lin, K. Price, G. B. Rayner, S. Khandelwal, S. Salahuddin, and A. D. Franklin. Sustained sub-60 mv/decade switching via the negative capacitance effect in mos2 transistors. Nano letters, 17(8):4801–4806, 2017.DOI
- [36]S. Mondal and V. Venkataraman. Low temperature below 200° c solution processed tunable flash memory device without tunneling and blocking layer. Nature Communications, 10(1):2143, 2019.DOI
- [40]M. H. Park, Y. H. Lee, H. J. Kim, Y. J. Kim, T. Moon, K. D. Kim, J. Mueller, A. Kersch, U. Schroeder, T. Mikolajick, et al. Ferroelectricity and antiferroelectricity of doped thin hfo2-based films. Advanced Materials, 27(11):1811–1831, 2015.DOI
- [43]P. Ravikumar, P. Venkatesan, C. Park, N. Afroze, M. Tian, W. Chern, S. Datta, S. Yu, S. Mahapatra, and A. Khan. The first switch effect in ferroelectric field-effect transistors. IEEE Transactions on Device and Materials Reliability, 2025.
- [44]P. G. Ravikumar, C. Park, P. V. Ravindran, N. Afroze, M. Tian, W. Chern, S. Datta, S. Yu, S. Mahapatra, and A. Khan. First write pulse-induced interface damage in ferroelectric field-effect transistors. In 2024 IEEE International Integrated Reliability Workshop (IIRW), pages 1–4. IEEE, 2024.
- [46]S. Salahuddin and S. Datta. Use of negative capacitance to provide voltage amplification for low power nanoscale devices. Nano letters, 8(2):405–410, 2008.
- [47]S. Sett, R. Debnath, A. Singha, S. Mandal, K. Jyothsna, M. Bhakar, K. Watanabe, T. Taniguchi, V. Raghunathan, G. Sheet, et al. Emergent inhomogeneity and nonlocality in a graphene field-effect transistor on a near-parallel moiré superlattice of transition metal dichalcogenides. Nano Letters, 24(30):9245–9252, 2024.DOI
- [48]P.-C. Shen, C. Lin, H. Wang, K. H. Teo, and J. Kong. Ferroelectric memory field-effect transistors using cvd monolayer mos2 as resistive switching channel. Applied Physics Letters, 116(3), 2020.DOI
- [49]M. Si, A. K. Saha, S. Gao, G. Qiu, J. Qin, Y. Duan, J. Jian, C. Niu, H. Wang, W. Wu, et al. A ferroelectric semiconductor field-effect transistor. Nature Electronics, 2(12):580–586, 2019.
- [52]M. M. Waldrop. The chips are down for moore’s law. Nature News, 530(7589):144, 2016.DOI
- [53]M. M. Waldrop. More than moore. Nature, 530(7589):144–148, 2016.
- [55]X. Wang, K. Yasuda, Y. Zhang, S. Liu, K. Watanabe, T. Taniguchi, J. Hone, L. Fu, and P. Jarillo-Herrero. Interfacial ferroelectricity in rhombohedral-stacked bilayer transition metal dichalcogenides. Nature nanotechnology, 17(4):367–371, 2022.DOI
- [57]A. Weston, E. G. Castanon, V. Enaldiev, F. Ferreira, S. Bhattacharjee, S. Xu, H. Corte-León, Z. Wu, N. Clark, A. Summerfield, et al. Interfacial ferroelectricity in marginally twisted 2d semiconductors. Nature nanotechnology, 17(4):390–395, 2022.DOI
- [58]Y. Wu, Z. Zeng, H. Lu, X. Han, C. Yang, N. Liu, X. Zhao, L. Qiao, W. Ji, R. Che, et al. Coexistence of ferroelectricity and antiferroelectricity in 2d van der waals multiferroic. Nature Communications, 15(1):8616, 2024.DOI
- [60]K. Yasuda, X. Wang, K. Watanabe, T. Taniguchi, and P. Jarillo-Herrero. Stacking-engineered ferroelectricity in bilayer boron nitride. Science, 372(6549):1458–1462, 2021.
- [61]J. Yu, H. Wang, F. Zhuge, Z. Chen, M. Hu, X. Xu, Y. He, Y. Ma, X. Miao, and T. Zhai. Simultaneously ultrafast and robust two-dimensional flash memory devices based on phase-engineered edge contacts. Nature Communications, 14(1):5662, 2023.
- [62]M. Yuan, B. Zhang, J. Cai, J. Zhang, Y. Lu, S. Qiao, K. Cao, H. Deng, and Q. Ji. In-plane ferroelectrics enabling reduced hysteresis in monolayer mos2 transistors. Carbon Neutralization, 3(4):700–709, 2024.
- [63]P. Yudin, M. Y. Hrebtov, A. Dejneka, and L. McGilly. Modeling the motion of ferroelectric domain walls with the classical stefan problem. Physical Review Applied, 13(1):014006, 2020.
- [66]P. Zhao, A. Khosravi, A. Azcatl, P. Bolshakov, G. Mirabelli, E. Caruso, C. L. Hinkle, P. K. Hurley, R. M. Wallace, and C. D. Young. Evaluation of border traps and interface traps in hfo2/mos2 gate stacks by capacitance–voltage analysis. 2D Materials, 5(3):031002, 2018.
- [67]C. Zhou and Y. Chai. Ferroelectric-gated two-dimensional-material-based electron devices. Advanced Electronic Materials, 3(4):1600400, 2017.DOI

