WorksPhysics and Astronomy
High-Performance Ferroelectric Field-Effect Transistors with Ultra-High Current and Carrier Densities
Abstract
Ferroelectric field-effect transistors (FeFET) with two-dimensional (2D) semiconductor channels are promising low-power, embedded non-volatile memory (NVM) candidates for next-generation in-memory computing. However, the performance of FeFETs can be limited by a charge imbalance between the ferroelectric layer and the channel, and for low-dimensional semiconductors, also by a high contact resistance between the metal electrodes and the channel. Here, we report a significant enhancement in performance of contact-engineered FeFETs with a 2D MoS2 channel and a ferroelectric Al0.68Sc0.32N (AlScN) gate dielectric. Replacing Ti with In contact electrodes results in a fivefold increase in on-state current (~120 uA/um at 1 V) and on-to-off ratio (~2*10^7) in the FeFETs. In addition, the high carrier concentration in the MoS2 channel during the on-state (> 10^14 cm^-2) facilitates the observation of a metal-to-insulator phase transition in monolayer MoS2 permitting observation of high field effect mobility (> 100 cm^2V^-1s^-1) at cryogenic temperatures. Our work and devices broaden the potential of FeFETs and provides a unique platform to implement high-carrier-density transport in a 2D channel.
Introduction
The ferroelectric field effect transistor (FeFET) device has garnered attention due to its desirable properties such as a compact, non-volatile memory, high on-to-off current ratio, high access speed, and low programming energy.[1] However, the majority of FeFETs employing conventional semiconductor channels, such as Si, face a significant challenge due to the charge imbalance between the ferroelectric layer and the channel.[2, 3] For example, when HfZrO, which shows a remnant polarization () of 10–30 μC/cm, is used in the gate stack, the carrier density in the Si channel should ideally reach 10 cm to achieve a charge balance with the of HfZrO. However, the carrier concentration in conventional Si channel FeFETs can only reach up to 1–2×10 cm due to charge screening by the native SiO interface and the 3D nature of the semiconductor.[2, 3] This results in a large discrepancy in charge density between the channel and HfZrO.[2] This mismatch causes partial polarization of the FeFETs and the accumulation of trapped carriers at the Si/HfZrO interface, resulting in various reliability issues.[2, 4, 5] On the other hand, several reports have highlighted the doping capacity of 2D materials, particularly MoS, in supporting carrier concentration exceeding 10–10 cm.[6, 7, 8, 9] This is particularly advantageous to mitigate the charge imbalance issue between the ferroelectric and 2D channel material. The substantial capability of 2D MoS to support higher carrier concentrations owing to its 2D density of states further makes them promising channel material candidates, especially when paired with ferroelectric materials with large in the gate stack.[8, 10, 11, 12, 13, 14, 15]
Recently, AlScN (AlScN) has been explored as the ferroelectric material in 2D FeFETs due to its high value (80–115 μC/cm) and the feasibility of using a low thermal budget process for back-end-of-line direct integration.[16, 17] The high of AlScN, coupled with the high doping capacity and ultraclean van der Waals interface of 2D MoS, can offer an excellent combination for high-performance FeFETs.[17, 18] Specifically, the carrier density induced by the of AlScN is theoretically 5–7×10 cm (=, where = 1.6 × 10 C is the elementary charge),[19] matching well with the doping capacity of MoS (10–10 cm).[6, 7, 8, 9] However, the reported MoS/AlScN FeFETs[17, 18] may not have fully taken advantage of the high of AlScN due to imperfect metal-semiconductor contacts.[20, 21, 22, 23, 24] For example, high contact resistance results in degradation in sheet conductance, transconductance, and on-state () and off-state currents (), ultimately impeding the performance of ultra-scaled 2D devices.[20, 21, 22, 23, 24] Therefore, to fully utilize the high of AlScN, it is important to tackle the persistent challenge of high contact resistance between the 2D channel and contact electrodes. This entails resolving issues such as energy level mismatch at the metal-semiconductor junction (MSJ) interface,[20, 21] high-energy metal deposition processes[22], and chemical reactions of metals with 2D semiconducting transition metal dichalcogenides (TMDs).[20, 24] Moreover, in the case of 2D FeFETs, there have been no systematic investigations into the effects of contact interface on key performance indicators such as , and programming/erasing voltage.[8] This is particularly significant because low conductance may result in low read current and slow write/erase speed at the same read voltage.[2, 4, 5, 8] Furthermore, ensuring scalable production of low-resistance contact electrodes is crucial for the high reproducibility and reliability of 2D-channel/ferroelectric FeFETs for eventual production and commercialization.
Here, we report the high current and high carrier density in In-contact based MoS channel FeFETs with AlScN as the ferroelectric. We observe the of In-contacted MoS/AlScN FeFETs approaching ∼300 (∼120) μA/μm at a drain-to-source voltage () of 3 (1) V, resulting in a high of ∼2 × 10 on average, which is ∼3 times higher than with Ti contacts. Remarkably, (∼100 μA/μm), (∼2 × 10), and memory windows normalized by the ferroelectric thickness (∼0.11 V/nm) of the MoS/AlScN FeFETs surpass those reported for any other 2D-channel/ferroelectric FeFETs. In conjunction with the negligible thermal emission barrier height of the In-MoS MSJ interface, the high of the AlScN enables a significantly large number of carriers to be injected into the monolayer channel (> 4.8 × 10 cm) with a high field-effect mobility (∼31 cmVs). In addition, the large doping capacity enables the observation of a metal-to-insulator phase transition in monolayer MoS, concurrently with a band-like transport as observed in variable temperature measurements further suggesting high doping levels enabled by the high of AlScN. Finally, we have also explored the pulse dynamic characteristics and successfully demonstrated the realization of multilevel conductance states in the FeFETs, which is important for amplifying effective bit density for embedded memory technologies. Notably, our FeFETs with In contact exhibits a larger read current of ∼3.1 (5.2) μA in response to a 0.1 (1.0) ms programming pulse compared to those using Ti contacts, suggesting potential for low-power operation.
RESULTS AND DISCUSSION
FeFETs with different contact interfaces.
Figure 1a illustrates the cross-sectional schematic of the FeFET structure. A MoS monolayer grown by metal-organic chemical vapor deposition (MOCVD) is utilized as the 2D channel and the AlScN layer, 45 nm thick, acts as an effective ferroelectric gate due to its high (80–115 ).[16, 17] The applied gate voltage () induces the switching of ferroelectric polarizations in AlScN, resulting in a change of the fermi level () of MoS. This modulation subsequently produces erased or programmed conductance states of the FeFET (Figure 1b). Here, metallic In has been chosen as contact electrodes to integrate with the MoS/AlScN structure with a low contact resistance. This is because In is known to be less chemically reactive with TMDs and has a work function ( eV) that is smaller than the electron affinity of monolayer MoS ( eV), possibly resulting in the a negligible Schottky barrier height at the MSJ interface (i.e., meV for ideal case).[20, 22] The device fabrication is scalable (see Methods for detail), which allows the successful fabrication of arrays of In-contacted MoS/AlScN FeFETs (Figure 1c). The device dimensions are a channel length () of 300 to 1,500 nm and a width () of 20 m, as presented in a scanning electron microscopy (SEM) image (Figure 1d).

Figure 1. Structure and electrical properties of a monolayer MoS/AlScN FeFET. (a) Schematic diagram of a MoS/AlScN FeFET with In/Au contact electrodes. (b) Illustrations depicting the band bending of monolayer MoS, influenced by the polarization directions ( or ) of AlScN, which in turn govern the erase/program functions of the FeFET. The polarization bound charges with + or - symbols and mobile electrons are displayed in the bands. (c, d) SEM images of (c) the FeFET arrays and (d) the unit device with a channel width () of 20 μm. (e, f) Transfer characteristics (-) of the FeFET with In (red) and Ti contact electrodes (blue) under the of 1 V in the (e) semi-logarithm and (f) linear scales. The transfer curves were recorded at a rate of 10 Hz with a spacing of 0.2 V. The differences in the subthreshold voltages () indicate the memory window (MW). (g) Output characteristic (-) of FeFET under the of 12 V (See Figure S3 for the dependence of output curves on .). (h-j) Histograms of (h) values showing the memory windows, (i) on-state current density () at = 10 V, and (j) / values depending on the In (red) and Ti contacts (blue). The performance metrics are extracted using the transfer curves under of 1 V for devices with a channel length () of 300-1,500 nm. / is the maximum attainable ratio between and .
The typical transfer curve of an In-contacted FeFET, measured at of 1 V in both semilogarithmic and linear scales, is shown as red curves in Figure 1e and Figure 1f, respectively. The transfer characteristic of the FeFET displays a counterclockwise sweep direction with high gate-field modulation () (Figure 1e), indicating the initiation of the ferroelectric switching mechanism. On the other hand, the MoS FETs fabricated with a non-ferroelectric (non-Fe) dielectric gate (SiO) shows clockwise hysteresis caused by charge trap effects, further confirming the effect of AlScN (Figure S1). The memory window (MW), extracted by the threshold voltages () shift, is confirmed to be greater than 7 V (Figure 1f). To prove the impact of In contacts, the FeFETs are also fabricated with Ti contacts (instead of In) for comparison, and the transfer curve results of the fabricated devices are displayed as blue curves in Figure 1e and f. Ti was selected for the comparison because it is chemically reactive when deposited on MoS, resulting in a degraded MSJ interface with high contact resistance.[20, 25] As proof, a representative Incontacted MoS/AlScN FeFET shows a higher of A/m than that of Ti-contacted ( A/m) at of 1 V. Such a high on-state current density in monolayer MoS can also be realized in an FeFET with thicker AlScN ( nm), indicating the universality of the In contact for high resistance ratio switching (Figure S2). The representative output characteristics also display times higher saturation current in the FeFETs with In contacts over Ti contacts (Figure 1g and Figure S3). The saturated of A/m is reached at only of V for the FeFET with a 300 nm channel length, which is much higher than that of a Ti/MoS/AlScN FeFET ( A/m). It is worth noting that the saturation of our In-contacted FeFETs ( A/m) surpasses the highest ( A/m) reported previously for Ti-contacted MoS/AlScN FeFETs with a 4-times shorter of nm.[18]

Figure S1. Ferroelectric switching mechanism of FeFET with In contacts. (a) Transfer characteristics for various sweep ranges (, , and V), all showing counterclockwise hysteresis, as indicated by the arrows. (b) Comparison of the - plots for FeFET (red) and standard FET (blue) with applied through AlScN (45 nm thickness) and SiO (50 nm thickness) gate substrates. Both devices were subjected to a sweep up to V for fair comparison. (c) Zoomed-in transfer curve of the FET gated with SiO non-FE dielectric, from panel (b). The arrows indicate the hysteresis directions.

Figure S2. High-performance of FeFET with a thicker AlScN ferroelectric. (a) Cross-sectional schematic of the device structure. (b) Transfer and (c) output curves of the 100 nm- and 45-nm thick AlScN-based FeFETs. The output curves are measured under the maximum in (b), which is 10 and 14 V for 45 and 100 nm AlScN case, respectively.

Figure S3. Gate-dependent output characteristics of FeFETs. The output curves (–) of MoS/AlScN FeFET with (a) In and (b) Ti contact electrodes. In contact FeFET shows more than five times higher than Ti contact FeFET when is 12 V.
To enhance the impact and reproducibility of our results, we investigate the performance of Ti and In contact electrodes on a large number of monolayer MoS/AlScN FeFETs, with over 20 devices tested for each type of contact rather than a comparison from a single representative I-V characteristic. The results are compiled and presented in the form of histograms (Figure 1h-Figure 1j). The results show that devices with In contacts have times higher ( A/m) and times better () than devices with Ti contacts, when averaging 20 devices with varied from to 1,500 nm (Figure 1i, Figure 1j). On the other hand, the MW of the MoS/AlScN FeFET with In contacts ( V) is smaller compared to the devices with Ti contacts ( V) (Figure 1h). The smaller MW is a result of the negative shift of in the forward sweep, which is caused by the increased efficiency of current injection by the small-work-function In contact that makes it easier for the transistor to switch on.[26, 27] The variability of with respect to is characterized by calculating the coefficient of variation (CV = , where is the standard deviation and the mean value) for the In- and Ti-contacted MoS/AlScN FeFETs. The average CV for the In-contacted FeFETs is , which is three times smaller than that for the Ti-contacted devices () for the devices with the same ( nm). The observed device-to-device variation is attributed to the non-uniformity of the contact and dielectric interface, as well as local charge inhomogeneities.[28]
Transmission line method for FeFETs.
To investigate the room-temperature electrical properties of the FeFETs, contact resistance () and channel’s sheet resistance () are extracted using the transmission line method (TLM). Figure 2a shows a plot of total resistance normalized by channel width () of the FeFETs with In and Ti contacts at of 10 V. The averaged increases linearly as a function of the according to a relationship as follows:

Figure 2. High resistance ratio switching in the FeFET with low contact resistance. (a) Total resistance () of the FeFETs with In and Ti contact electrodes depending on the channel length () measured at of 10 V during the forward sweep in transfer transport. The -intercepts and the slope indicate the contact resistance () and the channel sheet resistance () of the TLM sets, respectively. (b) and (c) of In- (red) and Ti-contacted FeFETs (blue) under the different in the forward (solid) and backward sweep (open). (d) Comparison of as a function of of few-layer FETs[21, 22, 23, 24, 29, 30] (circles) and our /AlScN FeFET (stars). The contact electrodes for each FET are labeled. IL stands for interlayer contact. (e, f) Comparisons of (e) (normalized at = 1 V) depending on the channel length () and (f) / and memory window (normalized by the thickness of ferroelectric) for our FeFET (red stars) with the literature values for FET[6, 20, 21, 22, 23, 36, 37, 38, 40] (open) and FeFET[10, 11, 31, 32, 33, 34, 35] (solid). Compared values are extracted from the DC measurement for the transfer characteristic of (Fe)FETs. The fitted curve in (e) reveals the ohmic relation of the values (). For standard (non-ferroelectric) FETs, the contact electrodes are labeled; and for FeFETs, the ferroelectric material is labeled. The reported performances for the Si-channel HZO FeFETs[41, 42] (purple squares) are also displayed in (f) to compare with 2D -based FeFETs.
Here, is field-effect mobility (=, where is the dielectric constant of [16, 17]). Using Equation (2), one can obtain $2R_{\mathrm{c}}$ and as they are the -intercept and the slope of the - plot, respectively (Figure 2a). The calculated on-state value of In/ MSJ FeFET approaches 4.5 1.9 k at of 10 V (average standard error from the linear fitting), which is almost two times smaller than that of Ti/ MSJ FeFET (8.8 0.9 k) at the same (Figure 2b), representing better MSJ interface with In contact. Moreover, the TLM-extracted of In-contacted FeFET decreases over , reaching 12.2 2.2 k/sq at 10 V (Figure 2c). The Ti/ MSJ FeFET, meanwhile, demonstrates a higher on-state (19.9 1.3$ k/sq). When comparing the smaller and values with those of FeFETs employing Ti contact electrodes, the effectiveness of carrier transport through the In contact is evident (More explanation on how leads to a smaller is provided in the following paragraphs for Figure 3.).

Figure 3. Extraction of mobility and carrier density during the ferroelectric gating. (a) Field-effect mobility () of the MoS/AlScN FeFETs with In (red) and Ti contact electrodes (blue) depending on the sweep directions. (b) Extracted carrier density () of the In-contact FeFET depending on . The calculated by the parallel plate capacitance model is displayed as solid gray lines. (c) Electrical polarization () calculated using the difference between values () of the data and the capacitance model in (b). (d) Comparison of the representative values of the In- and Ti-contacted FeFETs. The arrows indicate forward (dashed line) and backward (straight line) sweep directions during the transfer curve characterizations in Figure 1d, Figure 1e. The lines in (c) and (d) are guides for the eyes. (e-f) Comparisons of the FeFET and standard (non-Fe) FET in terms of (e) transfer characteristic during the backward sweep, and (f) TLM-extracted channel sheet conductance () as a function of . The standard FET has the same device configurations, dimensions, and materials as those in FeFET, except for the gate dielectric; i.e., SiO instead of AlScN. The of FeFET in (f) is calculated by assuming that the intrinsic field-effect mobility () calculated for the standard FET is comparable to that of FeFET.
To evaluate the performance of our In-contact FeFETs, we have conducted a comprehensive comparison with other devices reported, considering parameters such as , , , , and MW. Compared to the previous reports of FETs (with a non-Fe dielectric gate),[21, 22, 23, 24, 29, 30] the on-state of our FeFET is significantly smaller ( 11.4 k/sq), even smaller than the quantum resistance () (Figure 2d). Even though it has slightly higher (4.4–17 k) compared to the other studies on standard FETs[21, 22, 23, 24, 29, 30] (0.1–5 k), the is lower, indicating significant conductance switching in our FeFETs. Furthermore, our FeFET outperforms all the previously reported -based FeFETs,[10, 11, 31, 32, 33, 34, 35] as demonstrated by its improved (solid circles in Figure 2e), , and MW values (Figure 2f). In addition, the achieved high in our device (130 A/m) is higher than the previous report for FETs with In contact (18 A/m)[22] and 2D metal-based contact electrodes (0.1-3.4 A/m).[20, 36, 37] It is even higher than the FET with substitutionally doped channels (0.05-4 A/m)[6, 38] and comparable to FETs with semi-metal contacts of Sb (124-569 A/m)[23] or Bi (226 A/m)[21] that approach the quantum limit (see Table S1 for more comparisons). Besides, our device has the highest (2) and highest normalized MW (0.11 V/nm; normalized by the ferroelectric thickness) among -based FeFETs reported to date (Figure 2f),[10, 11, 31, 32, 33, 34, 35] which can provide a better read margin to minimize the error rate for memory applications.[39] The superior performance of our FeFET is not only due to the increased current injection efficiency by the In contact electrodes but also due to the high remnant polarization (80-115 C/cm) and large coercive field (2-4.5 MV/cm) of the AlScN ferroelectric itself.[17] Theoretically, the and directly relate with the carrier density induced in the channel, , and the MW ( thickness of the ferroelectric), respectively.[19]
| Device | Channel | Gate dielectric or ferroelectric | Contact metal or doping method | (kΩ·μm) | ( cm) | (nm) | (cmVs) at RT | (cmVs) at cryogenic | (μA·μm) at V | (kΩ/sq) | MW (V/nm) | ratio | Ref. |
| FeFET | CVD-MoS | AlScN | In/Au | 4.43 | 480 | 300 | 23.8 | 153 (at 30 K) | 130 (313 at V) | 12.2 | 0.11 | This study | |
| FET | CVD-MoS | SiO | Au + AlO doping | 0.48 | 20 | 180 | 33.5 (intrinsic) | N/A | 300 | 8.1 | N/A | 16 | |
| FET | CVD-MoS | SiO | PtTe | 168 | 9 | 1,500 | 9.7 | N/A | 3.4 | 36.1 | N/A | 17 | |
| FET | CVD-MoS | SiN | Bi | 0.123 | 20 | 200 | 55 | 120 (at 77 K) | 339 at V | 17.9 | N/A | 11 | |
| FET | CVD-MoS | SiO | In | 3.3 | 5 | 1,000 | 28 (at 280 K) | 93 (at 77 K) | 18 | 50.6 | N/A | 12 | |
| FET | CVD-MoS | HfO | Sb | 0.209 | 30 | 1,000 / 400 / 100 | 87 (at 320 K) | 188 (at 50 K) | 124 / 268 / 569 | 8 | N/A | 5 | |
| FET | CVD-MoS | HfO | Graphene | 59 | 16.57 | 7,000 | 17 | N/A | 0.1 | N/A | N/A | 18 | |
| FET | CVD-MoS | SiO | Graphene | 115 | 3 | 4,000 | 35 | N/A | 5 | N/A | N/A | 19 | |
| FET | CVD-MoS | SiO | Re doping | 26.25 | 6.5 | 10,500 | N/A | N/A | 0.05 | N/A | N/A | N/A | 20 |
| FET | CVD-MoS (monolayer) | SiO | UHV Au | 2.75 | 10 | 750 | 33 (intrinsic) | 136 (at 81 K, intrinsic) | 12 | 30 | N/A | N/A | 13 |
| FET | CVD-MoS | SiO | Fe doping | 0.67 | 76 | 900 | 54 | 94 (at 100 K) | 4 | N/A | N/A | 6 | |
| FET | Exfoliated MoS (monolayer) | SiO | Sn | 200 | 2 | N/A | 60 | 105 (at 1.4 K) | N/A | 60 | N/A | N/A | 21 |
| FET | CVD-MoS | SiO | CuS | 27.3 | N/A | 2000 | N/A | N/A | 1 | 545 | N/A | 22 | |
| FET | CVD-MoS | SiO | Ti/Au + Low-T growth | 4.9 | 6.1 | 100 | 33 | N/A | 140 | 35.3 | N/A | 23 | |
| FET | Exfoliated few-layer MoS | hBN/SiO | MoSe/Ti/Au | 1.9 | 7.9 | 400 | 50 | 160 (at 160 K) | 59 | 30 | N/A | 24 |
Table S1. Table S1. Comparisons of -type (FE)-FETs with 2D semiconductor channels. For FeFET with 2D channel, the 3D ferroelectric gate of 0.7Pb(MgNb)O-0.3PbTiO (PMN-PT), BiLaFeMnO (BLFMO), CuInPS (CIPS), poly(vinylidene fluoride-co-trifluoroethylene) (P(VDF-TrFE)), HfZrO (HZO) were used.
Contact-limited mobility and carrier density.
The current density of a transistor is directly related to mobility (or saturation velocity) and carrier density,[40] hence, to gain a better understanding of performance metrics in our devices, we examine them in this section. First, the two-terminal, field-effect linear mobility () values of our device with are extracted (Figure 3a). The In-contact FeFET exhibits higher than the Ti-contact device, attributed to its better contact interface and lower contact resistance. The degradation of a contact interface by Ti results in an increase in the contact resistivity () and the sheet resistance of MoS underlying contact metals (; note that this is different from channel resistance, ), as; ,[43] as depicted in Figure S4, Figure S4. On the other hand, soft metal In can minimize the degradation of upon metal deposition because the structure of the underlying 2D MoS can be preserved.[22, 44] Similarly, the contact interfacial issue may reduce the intrinsic mobility of MoS under the contact () by , where the is the sheet carrier density of the semiconductor underneath the contact.[43]

Figure S4. Contact interface effects on device performance. (a, b) Illustration depicting the relationship between contact resistivity () and sheet resistance of MoS underlying contact metals (), highlighting the degradation of the contact interface by Ti. (c, d) Schematic representations of MSJ barrier heights and width depending on the contact interface. The schematic in (c) illustrates the significantly small thermionic barrier height () at the In-MoS MSJ even at -10 V (~79 meV; erase state), possibly caused by the degenerately doping by the carrier transfer from In to MoS (). The diagram on the right in (c) depicts the above , leading to the dominance of field emission over thermionic emission through the MSJ, which is completely different from the Ti-MoS MSJ case in (d).
We have also assessed the -dependent carrier density () of In-MoS/AlScN FeFETs in the on state by utilizing the TLM-extracted and values (Figure 3b and Figure S5). During the forward sweep, in the 2D channel increases with . At the moment the sweep direction changes from forward to backward, the still increases due to the non-volatility of the ferroelectric polarization, reaching a maximum of , when the of 10 V gradually reduces to 9.4 V. As decreases further from 9.4 V, decreases exponentially toward the off state (e.g., at V). According to the standard parallel capacitance model (gray curves in Figure 3b), the -induced in the channel is much larger than what can be achieved solely with the AlScN dielectric gating (i.e., at ). This implies that the () of the ferroelectric AlScN plays a crucial role in the -induced electron density in the 2D channel.

Figure S5. TLM-extracted channel resistivity () of the MoS/AlScN FeFETs measured at V. The varied depending on (a) In and (b) Ti contact. The are extracted by excluding -dependent in their total , according to Equation (1) and Equation (2) in the main text.
To determine the 2D carrier density induced solely by the ferroelectric polarization (), we subtracted the obtained using the capacitance model (gray curves in Figure 3b) from the calculated total (data points in Figure 3b). We assume that the simplest case, wherein the increase in is related to the capacitive and ferroelectric responses,[45, 46, 47] such that the total . We then scale the resulting values to calculate the ferroelectric polarization, values of the FeFET, as (shown in Figure 3c). For example, the maximum value at of 9.4 V () corresponds to a value of . Remarkably, the calculated value ( at 9.4 V) due to ferroelectric polarization in our device is much closer to the reported of our AlScN[16, 17] (–135 ). This indicates that ferroelectric gating in the In-contacted MoS/AlScN FeFETs is effectively supported by the high characteristic of AlScN. In addition, the result proves the values are also highly dependent on the contact properties. Using Ti as the contact electrode, instead of In, results in a maximum value of only (Figure 3d). This suggests that the high of the MSJ interface hinders the utilization of the high inherent in AlScN[16, 17] (–135 ). When is high, the switching of the MFM capacitor under the contact may be imperfect, resulting in significantly lower partial polarization. The degraded MoS layer (beneath the contacts) can create multiple trap sites, which may result in a depolarization field that prevents full switching of AlScN.[35]
Given the huge mismatch of and carrier density in conventional 3D Si/HfZrO FeFETs, leading to substantial charge trapping at the Si/ferroelectric interface and reliability issues,[2] our findings suggest that a 2D channel FeFET with a superior contact interface can support the high carrier densities minimizing imbalance of the charges between channel and ferroelectric materials. The 2D MoS channel with minimal oxidation,[18] compared to the Si with native oxide,[2, 4, 5, 35] may enable the strong coupling between the ferroelectric polarization and the induced charge in the 2D channel due to the prevention of charge screening from the oxide layer. Moreover, the sub-1 nm thickness of 2D MoS allows minimal effective body capacitance compared to a 3D semiconductor channel, which can lead to a smaller depolarization field.[48, 49]
Therefore, combinations of high- AlScN with ohmic-contact 2D channel FeFETs can lead to significant conductance switching. To demonstrate this even more thoroughly, non-Fe FETs (i.e. standard FETs) with the exact same device configuration, fabrication process and material quality (including contact electrodes, channels, etc.), except for the gate dielectric (i.e., SiO for FET), are fabricated for a comparison with AlScN/MoS FeFETs. The representative transfer curves of the standard FET and FeFET are displayed in Figure 3e. Noticeably, the on-state current density of the FeFET is substantially superior to non-Fe FETs, more than an order of magnitude.
Moreover, one can determine the values in AlScN FeFETs by comparing the standard FETs’ channel conductance () and the “intrinsic” field-effect mobility (), extracted by the TLM method (Figure 3f and Figure S5). The on-state in the AlScN FeFET calculated in this way is , closely aligning with the value obtained in Figure 3b (), further proving its high doping density induced by the ferroelectric. Even in prior reports, such a high doping capacity in monolayer MoS () is challenging in traditional FETs with solid-state, non-ferroelectric gate dielectrics (Figure S6 and Table S2).

Figure S6. Benchmark plots of different doping methods and gate field effects on MoS. (a) Comparison of carrier densities () in MoS with different thicknesses, under different doping methods, or with gating methods. This highlights the challenge of achieving high doping levels () in a thinner MoS as reported in previous studies.[1, 2, 3, 4, 5, 6, 7, 8, 9, 10] (b) Charge density () vs. cryogenic mobility for different gate dielectric layers in transistors with a monolayer (1L) MoS channel.[5, 6, 8, 9, 11, 12, 13] Each point on the plot is labeled according to the specific gate dielectric type used in the measurements. The dielectric constants () for SiO, HfO, and AlScN are 3.9, 25, and 14, respectively.[14, 15] Note that the dual-gated[9] and defect-healed[8] MoS are mechanically exfoliated, while others are grown using CVD.[5, 6, 11, 12, 13]
| Doping strategy | Channel | Gate | Contact metal | Maximum ( cm) | (cmVs) at RT | (cmVs) at cryogenic | (A·m) at V | ratio | at RT (k/sq) | Scattering coefficient () | Ref. |
| AlScN gating | 1L-MoS (CVD) | AlScN | In/Au | 480 | 23.8 | 153 (at 30 K) | 130 (313 at V) | 11.3 (TLM) 0.36 ( nm) | 1.69 | This study | |
| Electrical double layer gating | 15 nm MoS | Ion gel | Au | 75 | N/A | 86 (at 220 K, Hall device) | 0.85 | 1.2 | N/A | 1 | |
| Electrical double layer gating | 20 nm MoS | Ion gel (DEME-TSFI) and HfO dual gate | N/A | 1.4 | N/A | 240 (at 20 K, Hall device) | N/A | N/A | 2 (at 100 K) | N/A | 2 |
| Alkali intercalation | Bulk MoS with K, Cs, Li | N/A | N/A | 400–1000 | N/A | N/A | N/A | N/A | N/A | N/A | 3, 4 |
| Contact engineering | 1L-MoS (CVD) | HfO | Sb | 30 | 87 (at 320 K) | 188 (at 50 K) | 569 | 8 | 1.8 | 5 | |
| Substitutional doping | Fe-doped MoS | SiO | Cr/Au | 76 | 54 | 94 (at 100 K) | 4 | N/A | N/A | 6 | |
| Surface charge transfer doping | Oleylamine-doped 70 nm MoS | SiO | Au | 19 | 25 | N/A | 11.8 | 2.9 | N/A | N/A | 7 |
| Defect healing | Thiol-treated 1L-MoS | SiO | Ti/Pd | 5.7 | 30 (81 with four terminal device) | 85 at 10 K (& 320 with four terminal) | N/A | N/A | 16.9 | 0.72 | 8 |
| Dual gate | 1L MoS | SiO and HfO | Au | 15 | N/A | 174 (at 4 K with four terminal device) | N/A | N/A | 4.3 | 1.4 | 9 |
| Solid superionic conductor gate | Few-layer MoS | LaF gating | Au | 41 | 60 | 532 (at 2 K) | N/A | – | 2 | N/A | 10 |
Table S2. Comparisons of doping strategies for 2D MoS.
Low-temperature measurements.
To further understand the transport mechanism, an MoS/AlScN FeFET with In contacts is additionally measured at various temperatures ( = 13–300 K). Figure 4a demonstrates the transfer curves measured under the same sweep (-10 to 10 V) at of 240–300 K. One can observe that the turn-on voltage for electron transport increases, and the subthreshold swing (SS) gets steeper at the higher . This is because the required to switch the polarizations in ferroelectric AlScN becomes higher at low , resulting in the suppression of both the ferroelectric negative capacitance effect and hole conductance as decreases. Conversely, in the range of 13–120 K, the SS are almost comparable to each other, and the decreases at higher (Figure 4b). In this regime, the FET is no longer turned on/off by the ferroelectric switching but controlled by the purely capacitive response from the non-Fe dielectric, which is also indicated in the - hysteresis directions depending on in Figure S7. Consequently, in the subthreshold regime, the suppression of thermionic emission of the electrons over the Schottky barrier at the MSJ becomes more noticeable as the decreases, resulting in the and SS changes.

Figure 4. Low-temperature measurements of FeFETs with its different gating mechanisms. (a, b) Transfer curves of the same FeFET measured at temperatures of (a) – K and (b) – K, during the backward sweep. Inset shows the zoomed-in plots in a linear scale near the range that shows metal-insulator electrical phase transition (MIT). (c) -dependent channel conductance () at the applied of (red) and V. The negative slope () indicates its metallic behavior whereas the positive slope () implies its insulating property. (d, e) Extracted depending on the applied , at (d) – K and (e) – K. The specific that enables the MIT where the is labeled as . (e) -dependent of three FeFET devices, with the fitted phonon scattering model (). (f, g) Extraction of the effective thermionic barrier height () at the metal-semiconductor junction between the MoS and In contact. (f) Arrhenius plot ( vs. ) to calculate the values The negative slopes indicate the absence of (), which is the case for measured at to V. (g) Extracted depending on the . The inset shows the zoomed-in curve for the range of to V, displaying the negative for V.
Figure 4c displays -dependent channel conductance () at of -10 V (blue) and 10 V (red). First, we can observe that the slope of the curve changes around 200 K to some extent, indicating the transition from the non-Fe to Fe gating mechanism of the transistor. Besides, at a of -10 V, the decreases when the decreases, as expected for temperature-dependent thermionic emission in insulators (, blue). Conversely, at a of 10 V, the improves when the declines (, red), probably due to the injected higher carriers, indicating its metallic phase. Even for monolayer MoS transistors with strongly suppressed carrier scattering, a high carrier density greater than cm (or ; which is the case in our device as shown in Figure 2d) in a 2D channel is possible, which leads to a critical point that shows the transition between the insulating phase to the metallic phase.[50, 51, 52]
Accordingly, this metal-to-insulator (MIT) behavior does provide a good indication of gating-induced doping capacity to reach cm (based on the theory[50, 51, 52]). We extract depending on at different regimes (Figure 4d), and here, the at which MIT occurs, or , is the point at which changes from negative to positive. Noticeably, the is 1 V at 240 K, whereas is 4 V for 120 K, suggesting that Fe gating can support larger than non-Fe gating and trigger band-like transport in the 2D channel.[50, 51, 52, 53] Note that the is ~1 V at K which is close to the voltage range where the reaches the ~ cm in Figure 3b, indicating their consistency.
The high -induced band-like transport is also supported by the observation of linear mobility evolution following the phonon-limited carrier transport model of with a for K (Figure 4e). This extracted value is comparable to the theoretical exponent for monolayer MoS () with the optical phonon scattering, not suppressed by other variations such as Schottky barriers.[50, 53] The consistency of -dependent has been verified across three different devices in Figure 4e. In spite of device-to-device variations, all devices display carrier transport characterized by the same exponent , maintaining consistency at -300 K, where ferroelectric gating affects the carrier injection. Given that this phonon-limited transport mechanism can be exclusively observed in MoS FETs with dual gates or with ultra-low contact resistance,[23, 50] our approach indicates that AlScN-based gating can provide a useful platform to enable high-mobility, high-carrier density 2D channel devices on a large scale. For example, over 100 cmVs is observed at low temperatures in two devices (i.e., ~114 and 116 cmVs for Device 1 and 3, respectively, at 13 K in Figure 4e. Refer to comparisons with previous reports in Figure S6 and Table S2). It is worth noting that the field-effect mobility stays high at large carrier densities suggesting the ability of ferroelectric polarization to effectively screen charged impurity scattering from defects and interfaces. On the other hand, below 120-200 K, the mobility starts to saturate mostly limited by Coulomb scattering of defects. This saturation is also largely affected by the absence of a contribution from the ferroelectric switching to the carrier transport.
The effectiveness of transport through the AlScN-based FeFET with a high-quality contact interface is further confirmed by its smaller effective thermionic barrier height () for electrons to overcome (Figure 4f,Figure 4g), extracted by the following equation:
Here, is the junction area, and is the effective Richardson–Boltzmann constant. Hence, the value of can be obtained from the slope of the linear fit to the Arrhenius plot ( vs. ) as depicted in Figure 4f. The calculated - plot suggests that is substantially small even at -10 V (~79 meV) (Figure 4g). This is even smaller than those of the In (~375 meV)[22] and Bi contact (~130 meV),[21] and comparable to the Sb contact (~53 meV)[23] for MoS FET in the completely off state. Moreover, at above -7 V (even in the subthreshold regime), the values are negative (Figure 4g and its inset), indicating the large field emission, rather than thermionic emission, through the MSJ (as depicted in Figure S4, Figure S4).
Pulse measurements.
Lastly, we evaluate the resistive switching and the retention of In-contacted MoS/AlScN FeFET, by applying a voltage pulse at the gate (). Figure 5a shows the time retention characteristics of binary states. Two distinct current states with a program/erase over at V and V are achieved after poling the polarization of AlScN up and down by a 1 s program pulse of ±12 V. It is found that varying the program pulse amplitude induces various , offering the possibility of multilevel storage in the FeFET, beyond simple binary states. Figure 5b (and Figure S8) shows the multilevel conductance stored after each pulse with different amplitude from -12 to 12 V (with 0.5 V differences) and a width of 1 ms. The different are recorded depending on the , and the programming of is possible under a bigger than 6.5 V. The stepwise program state is clearly distinct for a total of 58 levels stored at a V with increasing pulse numbers, promising its 5-bit operation.

Figure 5. Memory retention and programming speed characterizations at room temperature. (a) Retention properties of the MoS/AlScN FET with In contact electrodes, at the program and erase states in response to pulses with an amplitude of 12 V and width of 1 s. values were recorded at V and V. (b) Multiple states programmed/erased under the pulse train with different amplitudes from to 12 V with a pulse width () of 50 ms. were probed at V and V after applying for program/erase pulse. (c, d) Program state currents of the FeFETs with different contact electrodes, as a function of pulse amplitude. The programming was 1 ms and 0.1 ms in the case of (c) and (d), respectively.

Figure S8. Evaluation of reconfigurability in In-contacted MoS/AlScN FeFET. Multilevel conductance stored after each pulse, as depicted in (a, b), with varying amplitudes from −12 to 12 V (with 0.5 V differences) and a width of 50 ms. The read are depicted in (c, d) linear and (e, f) log scale. The recording of different depending on indicates the feasibility of programming under a greater than V. A stepwise program state is clearly discernible for a total of 58 levels stored at V with increasing pulse numbers, suggesting its 5-bit operation potential.
The programming speeds of FeFET are further estimated by measuring the pulse-width dependence of the . Figure 5c, Figure 5d illustrate the detailed program characteristics of the FET examined by varying pulse amplitude ( ranging from 4 to 12 V) and width ( of 0.1 and 1.0 ms). As the amplitude increases, the corresponding increase gradually as a response from the AlScN ferroelectric polarization switching, where is higher for In-MoS FeFETs than Ti-contacted ones in both cases contributed from its larger induced at the same . This suggests that In-contact FeFETs can use a lower pulse amplitude or smaller pulse width compared to Ti-FeFETs to complete the switching for the same program state, demonstrating the great potential of FeFETs for low-power and high-speed non-volatile memory applications.
Conclusion
In summary, the impacts of contact resistance on the monolayer FeFETs are investigated with In contact electrodes. High saturated current density ( at ), , and normalized MW are achieved in the FeFET. This achievement is attributed to the low contact resistance at the In/ MSJ interface down to , coupled with high remnant polarization, , of AlScN. Consequently, a significantly large carrier density approaching is obtained, corresponding to a partially switched value of . Furthermore, the high doping concentration induced by the ferroelectric allows for observation of nearly ideal phonon-limited transport (with an exponent of 1.7) with field-effect mobilities at cryogenic temperatures. In addition, pulse measurements conducted in FeFETs demonstrate the realization of 58 conductance levels. We have observed that the stored values in In-contacted FeFETs surpass those of Ti-contacted devices under the same program pulse amplitude and width. Therefore, our findings present a reproducible approach for achieving the high-current-density and high-mobility 2D channel AlScN FeFETs. Moreover, the low contact resistance and reconfigurable operations of the studied FeFETs make them suitable for integration into low-power electronics, offering potential additional functionalities.
Methods
Device fabrication.
with a thickness of 45 nm was deposited on pre-deposited Pt using a pulsed-DC reactive sputtering deposition system (Evatec, CLUSTERLINE 200 II pulsed DC PVD).[16, 17] The co-sputtering was conducted using separate 4-inch Al and Sc targets, at a chuck temperature of 350 °C, with 10 sccm of Ar gas flow and 25 sccm of gas flow under the constant pressure of bar, which is the same as that utilized in our previous reports for FeFETs.[15, 18] The preferentially oriented (111) Pt layer promoted the growth of AlScN with highly textured ferroelectricity along the [1] direction. MOCVD-grown monolayer thin film was then transferred from its growth substrate of sapphire to as-grown AlScN using KOH-based wet transfer. A polymeric supporting layer of polymethyl methacrylate (PMMA) was coated on the /sapphire and then floated onto 0.1 M KOH solution to detach from the sapphire substrate. The detached PMMA/ free-standing layer was rinsed in distilled ionized water several times to avoid any damage to AlScN from the KOH-based residues. After the transfer of PMMA/ to AlScN, the PMMA was removed by acetone for 10 min. The integration process for the monolayer /AlScN FeFETs necessitates the prevention of any high-energy process, thermal budget above 180 °C, and exposure to basic solution to ensure high-quality FeFETs. Source and drain contact of In/Au (10/40 nm) were defined through e-beam lithography (Elionix ELS-7500EX) followed by thermal evaporation (Kurt J. Lesker Nano-36). In particular, the deposition rate for In metal should be adjusted to below 0.03 Å/s to minimize any high-energy damage to the surface during the thermal evaporation and to pursue small contact resistance[22]. It should be noted that, in this work, the Ti- and In-contacted FeFET arrays were fabricated on the same AlScN substrate with the same MOCVD-grown film, thereby there were fewer variations in the materials and the channel/dielectric interface. The channel-defining process was conducted using the reactive ion etcher (RIE) with oxygen gas (Jupiter II RIE Plasma Etcher). The Pt beneath the AlScN was used as a gate contact, where the gate contact region was exposed via the wet etching of AlScN on the substrate by dipping it in the KOH solution after the fabrication of the In//AlScN/Pt heterostructure. No additional heat treatment was conducted after the device fabrication.
Device characterization.
Room-temperature electrical measurements were conducted in the air at ambient temperature in a Lakeshore probe station using a Keithley 4200A semiconductor analyzer. Low-temperature electrical measurements were carried out by a Keithley 4200A semiconductor parameter analyzer in a closed-cycle cryogenic probe station (Advanced Research Systems, Inc.) with a base pressure of ~ torr. All statistically estimated values in the paper are reported as "average standard deviation," unless otherwise specified. SEM was used to capture images at 5 keV with the backscattered detector (TESCAN S8000X).
ASSOCIATED CONTENT
Supporting information with Figure S1-8 and Table S1-2
AUTHOR INFORMATION
Author Contributions.
S. Song and K.-H. Kim contributed equally to this work.
S. S., K.-H. K. and D. J. conceived the experiments and wrote the manuscript with input from all authors. S. S., R. K., and M. D. conducted the low-temperature measurements. N. T., C. C., and J. M. R. grew the . R.K. and M.D. assisted with cryogenic transport measurements. J. Z., and R. H. O. conducted the AlScN deposition and characterization.
CONFLICT OF INTEREST
The authors declare no conflict of interest.
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.
ACKNOWLEDGMENT
D.J. and S.S. acknowledge primary support from the Office of Naval Research (ONR) Nanoscale Computing and Devices program (N00014-24-1-2131). D.J. and K.H-K acknowledge partial support from the Air Force Office of Scientific Research (AFOSR) GHz-THz program grant number FA9550-23-1-0391. S.S. acknowledges partial support for this work by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (Grant No. 2021R1A6A3A14038492). M.D. and R.K. acknowledge funding from DOE grant DE-SC0023224. A portion of the sample fabrication, assembly and characterization were carried out at the Singh Center for Nanotechnology at the University of Pennsylvania, which is supported by the National Science Foundation (NSF) National Nanotechnology Coordinated Infrastructure Program grant NNCI-1542153. The MOCVD MoS samples were grown in the 2D Crystal Consortium Materials Innovation Platform (2DCC-MIP) facility at Penn State which is supported by the National Science Foundation under cooperative agreement DMR-2039351.

Figure 7. Transition of FET behavior with various . The conductance switching by field effect transitions from Fe gating to gating by non-Fe dielectric capacitance, with corresponding changes in - hysteresis directions, depending on , i.e., counterclockwise direction at (a) 300 K, (b) 260 K, and (c) 200 K, and clockwise direction at (d) 160 K, (e) 120 K, (f) 13 K.
| Exfoliated few-layer MoS | LaF gating (solid superionic conductor) | N/A | 41 | N/A | 60 | 532 (at 2 K) | N/A | 2 | 10-10 | 10 | ||
| FeFET | Exfoliated InSe | CIPS | In/Au | N/A | 10,000 | 728 | N/A | 0.14 | 0.03 | 10 | 25 | |
| CVD-MoS | C-doped P(VDF-TrFE) | Cr/Au | 25,000 | 30.2 | N/A | 0.006 at V | 0.059 | 10 | 26 | |||
| Exfoliated MoS (13 nm) | AlO/BLFMo | Au | 20,000 | N/A | 3.3 | 0.060 | 10 | 27 | ||||
| Exfoliated MoS (6 nm) | AlO/HZO | Ti/Au | 2,000 | N/A | 30 at V | 0.0375 | 10 | 28 | ||||
| CVD-MoS | PMN-PT | Ni/Au | 5,000 | N/A | 2.08 | N/A | 0.0002 | 10 | 29 | |||
| Exfoliated MoS (7 nm) | CIPS | Ni | 1,000 | N/A | 0.19 at V | 0.015 | 10 | 30 | ||||
| Exfoliated MoSe | P(VDF-TrFE) | Ti/Au | 5,000 | 6.77 | N/A | 0.001 at V | 0.08 | 10 | 31 | |||
| Exfoliated MoS (4.62 nm) | P(VDF-TrFE) | Cr/Au | 3,300 | 95.6 | N/A | 2.5 at V | 0.08 | 10 | 32 | |||
| CVD-MoS | HZO | Ti/Au | 1,000 | N/A | 0.01 at V | 0.066 | 10 | 33 |
Table 1.
Supporting References
Zhang, Y., Ye, J., Matsuhashi, Y. & Iwasa, Y. Ambipolar MoS2 Thin Flake Transistors. Nano Letters 12, 1136-1140 (2012). https://doi.org:10.1021/nl2021575
Ye, J. T. et al. Superconducting Dome in a Gate-Tuned Band Insulator. Science 338, 1193-1196 (2012). https://doi.org:doi:10.1126/science.1228006
Woollam, J. A. & Somoano, R. B. Superconducting critical fields of alkali and alkaline-earth intercalates of . Physical Review B 13, 3843-3853 (1976). https://doi.org:10.1103/PhysRevB.13.3843
Somoano, R. B., Hadek, V. & Rembaum, A. Alkali metal intercalates of molybdenum disulfide. The Journal of Chemical Physics 58, 697-701 (1973). https://doi.org:10.1063/1.1679256
Li, W. et al. Approaching the quantum limit in two-dimensional semiconductor contacts. Nature 613, 274-279 (2023). https://doi.org:10.1038/s41586-022-05431-4
Li, H. et al. Reducing Contact Resistance and Boosting Device Performance of Monolayer MoS2 by In Situ Fe Doping. Advanced Materials 34, 2200885 (2022). https://doi.org:https://doi.org/10.1002/adma.202200885
Lockhart de la Rosa, C. J. et al. Molecular doping of MoS2 transistors by self-assembled oleylamine networks. Applied Physics Letters 109 (2016). https://doi.org:10.1063/1.4972781
Yu, Z. et al. Towards intrinsic charge transport in monolayer molybdenum disulfide by defect and interface engineering. Nature Communications 5, 5290 (2014). https://doi.org:10.1038/ncomms6290
Radisavljevic, B. & Kis, A. Mobility engineering and a metal–insulator transition in monolayer MoS2. Nature Materials 12, 815-820 (2013). https://doi.org:10.1038/nmat3687
Wu, C.-L. et al. Gate-Induced Metal–Insulator Transition in MoS2 by Solid Superionic Conductor LaF3. Nano Letters 18, 2387-2392 (2018). https://doi.org:10.1021/acs.nanolett.7b05377
Shen, P.-C. et al. Ultralow contact resistance between semimetal and monolayer semiconductors. Nature 593, 211-217 (2021). https://doi.org:10.1038/s41586-021-03472-9
Wang, Y. et al. Van der Waals contacts between three-dimensional metals and two-dimensional semiconductors. Nature 568, 70-74 (2019). https://doi.org:10.1038/s41586-019-1052-3
English, C. D., Shine, G., Dorgan, V. E., Saraswat, K. C. & Pop, E. Improved Contacts to MoS2 Transistors by Ultra-High Vacuum Metal Deposition. Nano Letters 16, 3824-3830 (2016). https://doi.org:10.1021/acs.nanolett.6b01309
References
- [1]Khan, A. I., Keshavarzi, A. & Datta, S. The future of ferroelectric field-effect transistor technology. Nature Electronics 3, 588-597 (2020). https://doi.org:10.1038/s41928-020-00492-7DOI
- [2]Toprasertpong, K., Takenaka, M. & Takagi, S. On the strong coupling of polarization and charge trapping in HfO2/Si-based ferroelectric field-effect transistors: overview of device operation and reliability. Applied Physics A 128 (2022). https://doi.org:10.1007/s00339-022-06212-6DOI
- [3]Kim, K.-H., Karpov, I., Olsson, R. H. & Jariwala, D. Wurtzite and fluorite ferroelectric materials for electronic memory. Nature Nanotechnology 18, 422-441 (2023). https://doi.org:10.1038/s41565-023-01361-yDOI
- [4]Ni, K. et al. Critical Role of Interlayer in Hf0.5Zr0.5O2 Ferroelectric FET Nonvolatile Memory Performance. IEEE Transactions on Electron Devices 65, 2461-2469 (2018). https://doi.org:10.1109/ted.2018.2829122DOI
- [5]Deng, S., Liu, Z., Li, X., Ma, T. P. & Ni, K. Guidelines for Ferroelectric FET Reliability Optimization: Charge Matching. IEEE Electron Device Letters 41, 1348-1351 (2020). https://doi.org:10.1109/led.2020.3011037DOI
- [6]Li, H. et al. Reducing Contact Resistance and Boosting Device Performance of Monolayer MoS2 by In Situ Fe Doping. Advanced Materials 34, 2200885 (2022). https://doi.org:https://doi.org/10.1002/adma.202200885
- [7]Xu, K. & Fullerton-Shirey, S. K. Electric-double-layer-gated transistors based on two-dimensional crystals: recent approaches and advances. Journal of Physics: Materials 3, 032001 (2020). https://doi.org:10.1088/2515-7639/ab8270DOI
- [8]Luo, Z.-D., Yang, M.-M., Liu, Y. & Alexe, M. Emerging Opportunities for 2D Semiconductor/Ferroelectric Transistor-Structure Devices. Advanced Materials 33, 2005620 (2021). https://doi.org:https://doi.org/10.1002/adma.202005620
- [9]Li, X. et al. Nonvolatile Electrical Valley Manipulation in WS2 by Ferroelectric Gating. ACS Nano 16, 20598-20606 (2022). https://doi.org:10.1021/acsnano.2c07469DOI
- [10]Li, C. et al. High-Performance C60 Coupled Ferroelectric Enhanced MoS2 Nonvolatile Memory. ACS Applied Materials & Interfaces 15, 16910-16917 (2023). https://doi.org:10.1021/acsami.3c02610DOI
- [11]Si, M., Liao, P.-Y., Qiu, G., Duan, Y. & Ye, P. D. Ferroelectric Field-Effect Transistors Based on MoS2 and CuInP2S6 Two-Dimensional van der Waals Heterostructure. ACS Nano 12, 6700-6705 (2018). https://doi.org:10.1021/acsnano.8b01810DOI
- [12]Tong, L. et al. 2D materials–based homogeneous transistor-memory architecture for neuromorphic hardware. Science 373, 1353-1358 (2021). https://doi.org:doi:10.1126/science.abg3161
- [13]Wang, X. et al. Van der Waals engineering of ferroelectric heterostructures for long-retention memory. Nature Communications 12, 1109 (2021). https://doi.org:10.1038/s41467-021-21320-2DOI
- [14]Luo, Z.-D. et al. Dual-Ferroelectric-Coupling-Engineered Two-Dimensional Transistors for Multifunctional In-Memory Computing. ACS Nano 16, 3362-3372 (2022). https://doi.org:10.1021/acsnano.2c00079DOI
- [15]Kim, K.-H. et al. Tuning Polarity in WSe2/AlScN FeFETs via Contact Engineering. ACS Nano 18, 4180-4188 (2024). https://doi.org:10.1021/acsnano.3c09279DOI
- [16]Liu, X. et al. Aluminum scandium nitride-based metal–ferroelectric–metal diode memory devices with high on/off ratios. Applied Physics Letters 118 (2021). https://doi.org:10.1063/5.0051940DOI
- [17]Liu, X. et al. Post-CMOS Compatible Aluminum Scandium Nitride/2D Channel Ferroelectric Field-Effect-Transistor Memory. Nano Letters 21, 3753-3761 (2021). https://doi.org:10.1021/acs.nanolett.0c05051DOI
- [18]Kim, K.-H. et al. Scalable CMOS back-end-of-line-compatible AlScN/two-dimensional channel ferroelectric field-effect transistors. Nature Nanotechnology (2023). https://doi.org:10.1038/s41565-023-01399-yDOI
- [19]Toprasertpong, K., Takenaka, M. & Takagi, S. Memory Window in Ferroelectric Field-Effect Transistors: Analytical Approach. IEEE Transactions on Electron Devices 69, 7113-7119 (2022). https://doi.org:10.1109/TED.2022.3215667DOI
- [20]Song, S. et al. Atomic transistors based on seamless lateral metal-semiconductor junctions with a sub-1-nm transfer length. Nature Communications 13, 4916 (2022). https://doi.org:10.1038/s41467-022-32582-9DOI
- [21]Shen, P.-C. et al. Ultralow contact resistance between semimetal and monolayer semiconductors. Nature 593, 211-217 (2021). https://doi.org:10.1038/s41586-021-03472-9DOI
- [22]Wang, Y. et al. Van der Waals contacts between three-dimensional metals and two-dimensional semiconductors. Nature 568, 70-74 (2019). https://doi.org:10.1038/s41586-019-1052-3DOI
- [23]Li, W. et al. Approaching the quantum limit in two-dimensional semiconductor contacts. Nature 613, 274-279 (2023). https://doi.org:10.1038/s41586-022-05431-4DOI
- [24]English, C. D., Shine, G., Dorgan, V. E., Saraswat, K. C. & Pop, E. Improved Contacts to MoS2 Transistors by Ultra-High Vacuum Metal Deposition. Nano Letters 16, 3824-3830 (2016). https://doi.org:10.1021/acs.nanolett.6b01309DOI
- [25]Song, S. et al. Wafer-scale production of patterned transition metal ditelluride layers for two-dimensional metal–semiconductor contacts at the Schottky–Mott limit. Nature Electronics 3, 207-215 (2020). https://doi.org:10.1038/s41928-020-0396-xDOI
- [26]Kong, L. et al. Wafer-scale and universal van der Waals metal semiconductor contact. Nature Communications 14, 1014 (2023). https://doi.org:10.1038/s41467-023-36715-6
- [27]Liu, G. et al. Graphene-assisted metal transfer printing for wafer-scale integration of metal electrodes and two-dimensional materials. Nature Electronics 5, 275-280 (2022). https://doi.org:10.1038/s41928-022-00764-4DOI
- [28]Liu, Y. et al. Promises and prospects of two-dimensional transistors. Nature 591, 43-53 (2021). https://doi.org:10.1038/s41586-021-03339-zDOI
- [29]Andrews, K., Bowman, A., Rijal, U., Chen, P.-Y. & Zhou, Z. Improved Contacts and Device Performance in MoS2 Transistors Using a 2D Semiconductor Interlayer. ACS Nano 14, 6232-6241 (2020). https://doi.org:10.1021/acsnano.0c02303DOI
- [30]Tang, A. et al. Toward Low-Temperature Solid-Source Synthesis of Monolayer MoS2. ACS Applied Materials & Interfaces 13, 41866-41874 (2021). https://doi.org:10.1021/acsami.1c06812DOI
- [31]Zhang, Q. et al. Tunable Multi-Bit Nonvolatile Memory Based on Ferroelectric Field-Effect Transistors. Advanced Electronic Materials 8, 2101189 (2022). https://doi.org:https://doi.org/10.1002/aelm.202101189
- [32]Zhang, S. et al. Low Voltage Operating 2D MoS2 Ferroelectric Memory Transistor with Hf1-xZr xO2 Gate Structure. Nanoscale Research Letters 15, 157 (2020). https://doi.org:10.1186/s11671-020-03384-z
- [33]Xu, L. et al. Ferroelectric-Modulated MoS2 Field-Effect Transistors as Multilevel Nonvolatile Memory. ACS Applied Materials & Interfaces 12, 44902-44911 (2020). https://doi.org:10.1021/acsami.0c09951DOI
- [34]Liu, L. et al. Electrical characterization of MoS2 field-effect transistors with different dielectric polymer gate. AIP Advances 7, 065121 (2017). https://doi.org:10.1063/1.4991843DOI
- [35]Huang, K. et al. Hf₀.₅Zr₀.₅O₂ Ferroelectric Embedded Dual-Gate MoS₂ Field Effect Transistors for Memory Merged Logic Applications. IEEE Electron Device Letters 41, 1600-1603 (2020). https://doi.org:10.1109/LED.2020.3019681DOI
- [36]Yu, L. et al. Graphene/MoS₂ Hybrid Technology for Large-Scale Two-Dimensional Electronics. Nano Letters 14, 3055-3063 (2014). https://doi.org:10.1021/nl404795z
- [37]Chee, S.-S. et al. Lowering the Schottky Barrier Height by Graphene/Ag Electrodes for High-Mobility MoS₂ Field-Effect Transistors. Advanced Materials 31, 1804422 (2019). https://doi.org:https://doi.org/10.1002/adma.201804422
- [38]Gao, J. et al. Transition-Metal Substitution Doping in Synthetic Atomically Thin Semiconductors. Advanced Materials 28, 9735-9743 (2016). https://doi.org:https://doi.org/10.1002/adma.201601104
- [39]Mulaosmanovic, H., Breyer, E. T., Mikolajick, T. & Slesazeck, S. Ferroelectric FETs With 20-nm-Thick HfO₂ Layer for Large Memory Window and High Performance. IEEE Transactions on Electron Devices 66, 3828-3833 (2019). https://doi.org:10.1109/TED.2019.2930749DOI
- [40]McClellan, C. J., Yalon, E., Smithe, K. K. H., Suryavanshi, S. V. & Pop, E. High Current Density in Monolayer MoS₂ Doped by AlOx. ACS Nano 15, 1587-1596 (2021). https://doi.org:10.1021/acsnano.0c09078DOI
- [41]Cai, Z., Toprasertpong, K., Liu, Z., Takenaka, M. & Takagi, S. Understanding HZO Thickness Scaling in Si FeFETs: Low Operating Voltage, Fast Wake-Up, and Suppressed Charge Trapping. IEEE Transactions on Electron Devices, 1-7 (2024). https://doi.org:10.1109/TED.2024.3386508DOI
- [42]Liao, C. Y. et al. Multibit Ferroelectric FET Based on Nonidentical Double HfZrO₂ for High-Density Nonvolatile Memory. IEEE Electron Device Letters 42, 617-620 (2021). https://doi.org:10.1109/LED.2021.3060589DOI
- [43]Schulman, D. S., Arnold, A. J. & Das, S. Contact engineering for 2D materials and devices. Chemical Society Reviews 47, 3037-3058 (2018). https://doi.org:10.1039/C7CS00828GDOI
- [44]Kim, B.-K. et al. Origins of genuine Ohmic van der Waals contact between indium and MoS₂. npj 2D Materials and Applications 5, 9 (2021). https://doi.org:10.1038/s41699-020-00191-zDOI
- [45]Kalbitz, R., Gerhard, R. & Taylor, D. M. Fixed negative interface charges compromise organic ferroelectric field-effect transistors. Organic Electronics 13, 875-884 (2012). https://doi.org:https://doi.org/10.1016/j.orgel.2012.01.034
- [46]Dickens, B., Balizer, E., DeReggi, A. S. & Roth, S. C. Hysteresis measurements of remanent polarization and coercive field in polymers. Journal of Applied Physics 72, 4258-4264 (1992). https://doi.org:10.1063/1.352213DOI
- [47]Liao, P. J. et al. in 2021 Symposium on VLSI Technology. 1-2.
- [48]Hoffman, J. et al. Ferroelectric Field Effect Transistors for Memory Applications. Advanced Materials 22, 2957-2961 (2010). https://doi.org:https://doi.org/10.1002/adma.200904327
- [49]Ma, T. P. & Jin-Ping, H. Why is nonvolatile ferroelectric memory field-effect transistor still elusive? IEEE Electron Device Letters 23, 386-388 (2002). https://doi.org:10.1109/LED.2002.1015207DOI
- [50]Radisavljevic, B. & Kis, A. Mobility engineering and a metal–insulator transition in monolayer MoS2. Nature Materials 12, 815-820 (2013). https://doi.org:10.1038/nmat3687DOI
- [51]Chen, X. et al. Probing the electron states and metal-insulator transition mechanisms in molybdenum disulphide vertical heterostructures. Nature Communications 6, 6088 (2015). https://doi.org:10.1038/ncomms7088DOI
- [52]Baugher, B. W. H., Churchill, H. O. H., Yang, Y. & Jarillo-Herrero, P. Intrinsic Electronic Transport Properties of High-Quality Monolayer and Bilayer MoS2. Nano Letters 13, 4212-4216 (2013). https://doi.org:10.1021/nl401916sDOI
- [53]Jariwala, D. et al. Band-like transport in high mobility unencapsulated single-layer MoS2 transistors. Applied Physics Letters 102 (2013). https://doi.org:10.1063/1.4803920DOI