Introduction

Sensitive, fast, and accurate detection of light is the foundation for the future optical communication, memory, sensing, imaging, and other optoelectronic applications. Photodetectors based on two-dimensional (2D) materials, such as graphene1,2 and transition metal dichalcogenides3,4,5, have emerged and drawn tremendous attention owing to their unique crystal structures, extraordinary electrical and optical properties, as well as the potential applications in ultrathin, transparent, and flexible optoelectronic devices6,7. The nature of a 2D optoelectronic material has largely determined the type and the working mechanism of its photodetectors8,9,10. Taking graphene as an example, as its gapless Dirac cone band structure results in ultrafast carrier recombination, graphene–metal junction and graphene p–n junction photodiodes are mainly exploited for separating photogenerated electrons and holes by the built-in electric field1,11,12,13. Graphene-based photodiodes with the characteristics of fast photoresponse but low photocurrent gain are hardly suitable for the room-temperature few-photon photodetection. On the contrary, molybdenum disulfide (MoS2) phototransistors are promising for visible to near-infrared highly sensitive photodetection, due to its thickness-dependent band gap, high photoconductive gain, and high carrier mobility14,15,16,17,18,19,20. Back in 2013, the monolayer MoS2 phototransistor was reported with responsivity (R) as high as 880 A W−1 at incident wavelength (λ) of 561 nm and power of 150 pW21. However, since its high responsivity is mainly attributed to trap states either in MoS2 or at the interface between MoS2 and SiO2 substrate, the uncontrollability of trap states caused by material and interface defects, low detectivity, and slow photoresponse on the timescale of seconds seriously limit its practical applications.

Nowadays, nearly all researches on improving the performance of MoS2 phototransistors focus on the hybrid systems combined with other material platforms such as graphene, quantum dots (QDs), perovskites, organic dye molecules, and surface plasmonic nanostructures22. For instance, the hybrid MoS2-PbS QDs photodetector was reported with an ultrahigh responsivity of 6 × 105 A W−1, yet large dark current of 0.26 μA, low detectivity, and high operating voltage23. Even though hybrid photodetectors using HgTe QDs demonstrate compelling sensitivity, the quality control of QDs and complicated preparation technology remain as the key challenges24. Besides, by incorporating organic dye molecules25 or perovskites26, MoS2 photodetectors demonstrate enhanced optical absorption and responsivity, whereas it is inevitable to change the characteristic absorption spectrum and form leakage paths, resulting in large dark current and low detectivity. Furthermore, using high-mobility graphene as an expressway for carrier transport of phototransistors can only improve the performance of MoS2 phototransistors to some extent, yet introducing undesirable complex interface and fabrication issues27.

Here, an alternative approach is developing gate dielectric engineering in MoS2 phototransistors. Ferroelectrics behave nonlinearly, which implies that phototransistors with a ferroelectric layer in the gate stack may have the ability to amplify the electrical response to optical signals. Ferroelectric hafnium zirconium oxide (HZO, Hf1−xZrxO2 with x = 0.5) thin film, with thickness scalability, complementary metal-oxide-semiconductor (CMOS) technology compatibility, and environmental friendliness, is thought as a kind of superior nonlinear gate dielectric layer to regulate optoelectronic characteristics of MoS2 phototransistors. Recent researches have demonstrated the photogating effect in MoS2 phototransistors, where trapped holes enable photogenerated electrons to circulate through an external circuit more than once leading to large photocurrent20,28. Besides, photogating effect can not only achieve the photoconductive gain but also lead to the threshold shift of phototransistors. Although the uncompensated dangling bonds or water molecules at the MoS2/oxide interface contribute to trap states to a certain extent, the ferroelectric HZO gate dielectric layer dominates and enhances the photogating effect by trapping more holes in the ferroelectric local electrostatic field, yielding high photoconductive gain and large threshold shifts. Moreover, ferroelectric negative capacitance (NC) effect, that the capacitance of ferroelectrics is negative during the process of ferroelectric polarization switching, possesses the ability to accomplish voltage amplification and ultra-steep subthreshold slope (SS) in the phototransistor with a ferroelectric layer in its gate stack29,30,31,32,33,34. Therefore, MoS2 phototransistors based on ferroelectric HZO films can efficiently amplify the change of the channel energy barrier induced by threshold shifts, resulting in that more electrons can go across the channel energy barrier and form large photocurrent. On the whole, ferroelectric HZO film is supposed to improve the performance of MoS2 phototransistors through strengthening the photogating effect by trapping more holes in the ferroelectric local electrostatic field and amplifying the energy barrier drop caused by threshold shifts based on ferroelectric NC effect.

In this paper, we report an ultrasensitive NC MoS2 phototransistor based on ferroelectric HZO thin film. In contrast to previously reported optimization strategies of MoS2 photodetectors, ferroelectric HZO thin films in the gate dielectric stack can significantly enhance the photogating effect, suppress the dark current and improve the light to dark current ratio by ferroelectric local electrostatic field and ferroelectric NC effect, thereby yielding an ultrahigh photodetectivity while maintaining high responsivity and fast response. Experimentally, by the gate structure engineering and the semiconductor thickness scaling35, MoS2 phototransistors accomplish a hysteresis-free ultra-steep SS (SSavg = 45 mV/dec, SSmin = 17.64 mV/dec). When illuminated, transfer characteristic curves shift to the left because of the strong photogating effect induced by ferroelectric local electrostatic field, yielding a change in phototransistor threshold voltage VT (in the dark) to VT − ΔVT (under illumination). As a result of threshold shift, the drain current (Ids) changes from Idark to Ilight = Idark + Iph and follows Iph = Ilight − Idark = ∫(dIds/dVg)dVT, where Iph is photocurrent, Idark and Ilight are the drain currents in the dark or under illumination. More specially, as dIds/dVg is fully dependent on SS in the subthreshold region, ultra-steep SS induced by ferroelectric NC effect can efficiently increase photocurrent and improve detectivity20. Therefore, based on the combined action of the threshold shift and the ultra-steep SS, we demonstrate the prototype devices with ultrahigh detectivity of 4.7 × 1014 cm Hz1/2 W−1 and high responsivity of 96.8 A W−1 under low operating voltages of Vds = 0.5 V and Vg = −1.6 V at room temperature. These results address the key challenges for ultrasensitive MoS2 photodetectors and are promising for the development of other photodetectors based on the photogating effect.

Results

Device structure and microscopic representation

Figure 1a, b is the structure schematic and the optical microscope photograph of the NC MoS2 phototransistor, respectively. A metal-ferroelectric-insulator-semiconductor field effect transistor (MFISFET) was fabricated on a p-type silicon substrate, with TiN as the gate metal, HZO/Al2O3 as ferroelectric/insulator gate layers, multilayer MoS2 as the channel semiconductor, and Cr/Au as the source and drain electrodes. The TiN layer is not only used as the back gate electrode but also as the substrate for growing high-quality ferroelectric HZO thin film by atomic layer deposition (ALD) process. The growth process of the ferroelectric HZO thin film is provided in “Methods” section, and the grazing-incident X-ray diffraction (GI-XRD) pattern of the ferroelectric HZO film is shown in Supplementary Fig. 1. Compared with conventional ferroelectrics, ferroelectric HZO thin film with thickness scalability, CMOS compatibility, and environmental friendliness is more applicable in future nanoscale electronic and optoelectronic devices. Besides, the amorphous Al2O3 layer plays an essential role in improving the quality of HZO/MoS2 interface36, including suppressing gate leakage current, achieving capacitance matching, and enhancing device stability31,32. As shown in Supplementary Fig. 2, metal-ferroelectric-semiconductor field effect transistors (MFSFETs) without the Al2O3 layer present inferior electrical properties. The inset in Fig. 1b is the atomic force microscope (AFM) imaging of a multilayer MoS2 flake, and the thickness of MoS2 is measured as 6.3 nm (nine layers) with the Raman spectrum of nine-layer MoS2 shown in Supplementary Fig. 337. Figure 1c, d is the cross-sectional transmission electron microscopy (TEM) imaging and the detailed energy dispersive X-ray spectrometry (EDS) elemental mapping of the channel and gate of the device. According to the TEM imaging, the thicknesses of MoS2/Al2O3/HZO/TiN layers are 6.3/6/10/40 nm, and the interfaces of all layers are clear, flat, and free of impurities. Besides, the layered structure of the MoS2 layer is discerned in Supplementary Fig. 4. Furthermore, the EDS analysis of Mo, S, Al, Hf, Zr, and Ti elements demonstrates that all elements are uniformly distributed and without inter-diffused.

Fig. 1: Schematic of device structure and microscopic representation.
figure 1

a Structure schematic of MoS2 phototransistors with the ferroelectric HZO thin film in the gate dielectric stack. b Optical microscope photograph of the device with MoS2 channel length of 3 μm, width of 15 μm. Scale bar, 5 μm. The inset is the atomic force microscope (AFM) imaging of the multilayer MoS2 flake, and the thickness of MoS2 is measured as 6.3 nm (9 layers). c Transmission electron microscopy (TEM) imaging of the channel and gate of the device is in order with MoS2/Al2O3/HZO/TiN layers. Interfaces of all layers are clear and flat. Scale bar, 10 nm. d Energy dispersive X-ray spectroscopy (EDS) elemental mapping of Mo, S, Al, Hf, Zr and Ti, corresponding to MoS2/Al2O3/HZO/TiN layers in the TEM imaging.

Electrical properties of NC MoS2 phototransistors

The ferroelectric properties of the HZO thin film, such as the hysteresis loop and the NC effect, play decisive roles in electrical and optoelectronic characteristics of NC phototransistors. Firstly, the solid blue line in Fig. 2a presents the hysteresis loop of the 10 nm-thick ferroelectric HZO thin film tested by a capacitor composed of Au/HZO/TiN layers (the maximum swept voltage of the P-E measurement is ±2.8 V, the frequency used is 1000 Hz, and the preset delay is 1000 ms). The hysteresis loop confirms the basic ferroelectric characteristics of HZO films with coercive field (Ec) of 1.5 MV cm−1 and remnant polarization (Pr) of 10 μC cm−2, and the dashed red line fitted by the Landau–Khalatnikov (LK) equation E = 2αP + 4βP3 + 6γP5 shows the negative dP/dE29,32. Furthermore, the ferroelectric capacitance (CFe) is defined as the reciprocal of the second derivation of the potential energy (U) with respect to the charge (Q), written as CFe = (d2U/dQ2)−1. The principle of the ferroelectric NC effect can be understood through device working mechanism schematics and ferroelectric double well energy landscapes34 shown in Fig. 2b. At the beginning, the gate voltage of the device is negative and less than the threshold voltage, yielding the ferroelectric layer polarized downwards, the channel electrons depleted, and the device turned off. Then changing the gate voltage to the opposite, the ferroelectric layer is polarized upwards, the channel electrons are accumulated, and the device is turned on. In the process of mutual conversion of the above two ferroelectric polarization states, the ferroelectric is supposed to go across an energy barrier from one energy valley to another. According to the definition of the ferroelectric capacitance, this energy barrier is the ferroelectric NC region.

Fig. 2: Electrical measurement and results of the NC MoS2 phototransistors.
figure 2

a Hysteresis loop (solid blue line) of the 10 nm-thick ferroelectric HZO thin film is measured by a capacitor composed of Au/HZO/TiN layers. The dashed red line fitted by the Landau–Khalatnikov (LK) equation E = 2αP + 4βP3 + 6γP5 shows negative dP/dE region. b Device working schematics and ferroelectric double well energy landscapes. During the process of ferroelectric polarization switching, the ferroelectric material is supposed to go across an energy barrier from one energy valley to another. This energy barrier is regarded as the ferroelectric NC region. c Equivalent capacitance schematic of the device, where CS, Cox and CFe represent the capacitance of MoS2, Al2O3 and HZO layers, respectively. d Output characteristic (VdsIds) curves exhibit ohmic characteristics for all gate voltages. e Forward and reverse transfer characteristic (VgIds) curves under Vds = 0.1 V are plotted in red and blue, respectively, and hysteresis between them is almost negligible. Electron mobility and SS are, respectively, extracted based on a linear scale (right axis) and a logarithmic scale (left axis). f SS is calculated according to the transfer characteristic curves, where the average value, the minimum values in forward and reverse (SSavg, SSmin-for, and SSmin-rev) are 45, 17.64, and 25.55 mV/dec, respectively.

To get insight into the fundamental electrical characteristics of the NC MoS2 phototransistors, Fig. 2c exhibits its equivalent capacitance schematic, where CS, Cox, and CFe represent the capacitance of MoS2, Al2O3, and HZO layers, respectively. Notably, CS also includes the stray capacitances of the source and drain shown in Supplementary Fig. 5. Cox38 and CFe can be calculated by the expression C = εSd−1, where ε, S, and d stand for the permittivity, area and thickness of dielectric layers, and the permittivity of the ferroelectric HZO thin film (εFe) presented in Supplementary Fig. 6 is measured by the capacitor shown in Fig. 2a. Through a simple calculation, the relationship between SS and capacitances is obtained as SS = [1 + CS(CFe−1 + Cox−1)] × 60 mV/dec = [1 − CS(|CFe|−1 − Cox−1)] × 60 mV/dec. According to this expression, it is beyond doubt that capacitance matching is crucial to achieve sub-60 mV/dec ultra-steep SS. Based on an optimized 10 nm-thick HZO ferroelectric layer and an appropriate 6 nm-thick Al2O3 layer (according to the expression of SS, |CFe|−1 − Cox−1 > 0 is a necessary condition for the sub-60 mV/dec ultra-steep SS), a series of thicknesses of MoS2 are measured experimentally and the measurement results are shown in Supplementary Fig. 7. As the thickness of MoS2 increases, that is the permittivity and capacitance of MoS2 increase35, the SS becomes steeper in accordance with the expression SS = [1 − CS(|CFe|−1 − Cox−1)] × 60 mV/dec. Meanwhile, when the thickness of the MoS2 is increased to 9.8 nm, a hysteresis about 0.16 V between forward and reverse VgIds curves is observed, which can be explained by the structural defects of MoS2 or the stability condition of NCFETs. For one thing, the total number of structural defects increases with the thickness of MoS239,40. When Vg sweeps forward, charge traps in the MoS2 layer inhibit the ferroelectric polarization switching by screening the electric field. After the traps have been neutralized, the ferroelectric polarization switches, leading to a large hysteresis and an abrupt jump in the reverse. For another, the stability condition of MFISFETs requires that the total gate-channel capacitance Cgc = (CS−1 + Cox−1 + CFe−1)−1 = (CS−1 + Cox−1 − |CFe|−1)−1 has to be positive29,31. Increasing the thickness (capacitance) of MoS2 may break the stability condition of the device, resulting in a large hysteresis in IdsVg curves.

Through optimizing the thickness of MoS2 in MFISFETs to 6.3 nm (nine layers), the MoS2 phototransistor accomplishes hysteresis-free ultra-steep SS. The output characteristic (VdsIds) curves shown in Fig. 2d exhibit ohmic characteristics for all gate voltages. The output characteristic curves under larger drain voltages are presented in Supplementary Fig. 8. In Fig. 2e, the forward and reverse transfer characteristic (VgIds) curves (the swept voltage of VgIds measurements is ramped and the swept rate is 14.14 Hz corresponding to the number of data points per second) are plotted in red and blue, respectively, and the hysteresis between them is almost negligible (the gate leakage measurement is shown in Supplementary Fig. 9). For a linear scale (right axis), electron mobility (μn) is calculated as 54.2 cm2 V−1 s−1 using the expression μn = (L/W)(S/Cins)Vds−1(dIds/dVg), where Cins stands for the total capacitance of gate insulators. For a logarithmic scale (left axis), SS calculated by the definition SS = dVg/d(lgIds) are presented in Fig. 2f, where the average value, the minimum value in forward and reverse (SSavg, SSmin-for, and SSmin-rev) are 45, 17.64, and 25.55 mV/dec, respectively. The optimization of electrical characteristics paves the way for ultrasensitive photodetection of MoS2 phototransistors working in the subthreshold region.

Mechanism of optical detection in NC MoS2 phototransistors

For typical photodetection operation, the laser is incident perpendicularly to the device surface, and the spot covers the entire channel uniformly shown in Fig. 3a. The transfer characteristics of the phototransistors in the dark and under the illumination of various optical powers are presented in Fig. 3b. The changes in threshold voltage VT,dark (in the dark) to VT,light = VT − ΔVT (under illumination) are observed in VgIds curves, and the threshold shift (ΔVT) increases with the incident optical powers. As a result of the threshold shift, the channel current increases from Idark to Ilight = Idark + Iph, that is the photocurrent Iph = Ilight − Idark = ∫(dIds/dVg)dVT, where dIds/dVg is fully dependent on SS in the subthreshold region. According to the definition of SS, with simple algebra the quantitative relationship among ΔVT, Iph and SS is obtained as lg(Ilight/Idark) = lgIlight − lgIdark = ∫SS−1dVT, indicating that the large threshold shift and the ultra-steep SS work together to contribute to the ultrahigh light to dark current ratio and achieve ultrasensitive photodetection. Dependence of threshold shifts on incident optical powers and dependence of photocurrent on threshold shifts are extracted from VgIds curves and presented in Fig. 3c. Since the ability of the ferroelectric local electrostatic field to trap holes is gradually weakened as holes accumulated more and more in the gate stack, P–ΔVT and ΔVTIph curves rise steeply for the incident optical power up to 0.2 nW and the threshold shift up to 0.21 V, respectively, and then gradually saturates.

Fig. 3: Mechanism of ultrasensitive optical detection in the NC MoS2 phototransistors.
figure 3

a Laser is incident perpendicularly to the device surface, and the spot covers the entire channel uniformly. b VgIds curves at Vds = 0.5 V in the dark and under various incident light powers (λ = 520 nm). c Dependence of threshold shifts on the incident optical powers and dependence of photocurrent on threshold shifts are extracted from VgIds curves. d Under light illumination, enhanced light absorption can be achieved in multilayer MoS2, generating electron-hole pairs efficiently by a strong light-matter interaction. e Photoexcited electrons are pulled toward the electrodes, while photoexcited holes are trapped in the gate stack due to the ferroelectric electrostatic local field, leading to strong photogating effect on the phototransistors. Photogating effect gives rise to a threshold shift as well as a reduction of the energy barrier in the channel (qΔV), so that more electrons can get over the energy barrier and forms photocurrent. f In generally, qΔV=qΔVT(1 + CSCins−1)−1 < qΔVT since the total capacitance of gate insulators (Cins) is usually positive. However, MoS2 NC phototransistors can break through this limitation and achieve a larger qΔV=qΔVT[1 − CS(|CFe|−1 − Cox−1)]−1 > qΔVT based on the ferroelectric NC effect and capacitance matching conditions.

To explore the mechanism of ultrasensitive photodetection, we focus on the working principles of optical absorption, photogating effect and voltage amplification in the device. Throughout the entire photodetection process, the operating voltages Vds and Vg are set as constant, and the gate voltage equals to the threshold voltage of the device in the dark. For the dark state, the ferroelectric layer is polarized by the gate voltage, and the channel electrons are depleted by the ferroelectric electrostatic local field, therefore the phototransistor is turned off, and the dark current is suppressed to an extremely low level. Under light illumination, the enhanced light absorption can be achieved in multilayer MoS2, generating electron-hole pairs efficiently by a strong light-matter interaction shown in Fig. 3d. Crucially, photoexcited electrons are pulled toward the external circuits, while photoexcited holes are trapped in the gate stack due to the ferroelectric electrostatic local field (Eferro), leading to a strong photogating effect on the MoS2 phototransistors. Furthermore, as presented in Fig. 3e, the photogating effect gives rise to a change of threshold voltage (ΔVT) as well as a reduction of the energy barrier (qΔV) in the channel, so that more electrons can get over the energy barrier and form photocurrent. As shown in Fig. 3f, the channel energy barrier drop qΔV=qΔVT(1 + CSCins−1)−1 is generally less than qΔVT, since the total capacitance of gate insulators is usually positive. However, MoS2 NC phototransistors can break through this limitation and achieve a larger qΔV=qΔVT[1 − CS(|CFe|−1 − Cox−1)]−1 > qΔVT based on the ferroelectric NC effect and capacitance matching conditions. It means that ultrasensitive photodetection of NC MoS2 phototransistors is fundamentally based on the combined action of the ferroelectric electrostatic local field and the ferroelectric NC effect.

Optoelectronic properties of NC MoS2 phototransistors

To confirm the above mechanism, we next turn our attention to the photoresponse of the NC MoS2 phototransistors. The optoelectronic measurements are performed under the illumination of a monochromatic light source with wavelength of 637 nm in a dark room at room temperature. Figure 4a presents the VdsIds curves in the dark and under various incident optical powers. To further evaluate the performance of the phototransistors, we extract the power dependence of photocurrent, light to dark current ratio, responsivity, external quantum efficiency (EQE) and detectivity from the VdsIds curves at Vds = 0.5 V and Vg = −1.6 V. With the logarithmic scale, the photocurrent dotted in red and the light to dark current ratio dotted in blue increase with the incident optical powers and achieve maximum values as high as Iph = 5.22 μA and Ilight/Idark = 5.8 × 107 at Peff = 1116 nW as shown in Fig. 4b. It is worth mentioning that Peff is calculated by Peff = APπ−1r−2, where A is the area of the sample, P is the actual laser output and r is the radius of the laser spot size. The responsivity R = IphPeff−1 indicates how much photocurrent can be produced by unit incident power on the photodetector, while the EQE defined as EQE = hcRλ−1e−1, is the ratio of the number of photoexcited carriers to the number of incident photons, where h is Planck constant, c is the speed of light and e is the electron charge. The responsivity and the EQE respectively plotted in Fig. 4c, d show the same changing tendency with the incident powers, reaching the largest values of R= 96.8 A W−1 and EQE = 1.88 × 104 % at the lowest incident power of Peff = 2.7 nW. The detectivity given by D* = RA1/2(2eIdark)−1/2 is one of the most important parameters for photodetectors. The ultrahigh detectivity of D* = 4.75 × 1014 cm Hz1/2 W−1 is achieved at Peff = 2.7 nW in Fig. 4e owing to the ultrahigh light to dark current ratio. Moreover, rise (τr) and decay (τf) times of photocurrent are measured as 400 μs and 200 ms, respectively, and both the rise and decay curves (in black) are fitted well using exponential functions (in red) shown in Fig. 4f.

Fig. 4: Optoelectronic measurement and results of the NC MoS2 phototransistors.
figure 4

a VdsIds curves in the dark and under various incident optical powers (λ  = 637 nm). b With the logarithmic scale, the photocurrent (in red) and the light to dark current ratio (in blue) increase with the incident powers and achieve maximum values as high as Iph = 5.22 μA and Ilight/Idark = 5.8 × 107 at Peff = 1116 nW. c, d, e Responsivity R = IphPeff−1, external quantum efficiency EQE = hcRλ−1e−1 and detectivity D* = RA1/2(2eIdark)−1/2 show the same changing tendency with the incident powers, reaching the largest values of R= 96.8 A W−1, EQE = 1.88 × 104 %, D* = 4.75 × 1014 cm Hz1/2 W−1 at the lowest incident power of Peff = 2.7 nW. f Rise (τr) and decay (τf) times of photocurrent are measured as 400 μs and 200 ms, respectively, and both the rise and decay curves (in black) are fitted well using exponential functions (in red).

We have also reviewed methods to improve the performance of MoS2 phototransistors based on previously reported researches and this work. As one kind of the most investigated 2D material-based photodetectors, optimization strategies and detailed performance parameters of MoS2 phototransistors are listed in Supplementary Table 1 to provide ideas for further exploration of other 2D optoelectronic semiconductors and their photodetectors41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62. There are four kinds of strategies regarding methods, including surface plasmon enhancement, charge-transfer assistance, impurity/energy band engineering of MoS2, and gate engineering. Among these methods, charge-transfer-assisted hybrid systems combined with other material platforms such as graphene, QDs, perovskites and organic molecules are the most extensively studied strategy recently. The hybrid phototransistors present enhanced optical absorption and responsivity, whereas it is usually inevitable to introduce leakage paths, increase dark current, and slower response speed. As presented in Fig. 5, NC MoS2 phototransistors based on gate dielectric engineering can circumvent these issues and demonstrate compelling performance with ultrahigh light to dark current ratio and detectivity, high responsivity and fast response, are promising for low-power and high-performance optoelectronic applications.

Fig. 5: Optimization strategy and performance summary of reported MoS2 phototransistors.
figure 5

Optoelectronic properties (a response time and the ratio of photocurrent to dark current. b responsivity and detectivity) of NC MoS2 phototransistors based on ferroelectric HZO gate dielectric engineering in this work outperform most of the previously reported MoS2 phototransistors based on other strategies regarding methods, including surface plasmon enhancement, charge-transfer assistance and impurity/energy band engineering of MoS2.

Discussion

In summary, based on the strategy of gate dielectric engineering, NC MoS2 phototransistors with HZO ferroelectric thin films in the gate dielectric stack are proposed and fabricated. The prototype photodetectors demonstrate ultra-steep subthreshold slope of 17.64 mV/dec and ultrahigh detectivity of 4.75 × 1014 cm Hz1/2 W−1 under λ = 637 nm and Peff = 2.7 nW at room temperature, owing to the strong photogating effect of ferroelectric local electrostatic field and the ultra-steep SS induced by ferroelectric NC effect during the process of ferroelectric polarization switching. This high-performance MoS2 photodetector should be a promising candidate for future optoelectronic applications and offer inspiration for the development of other optoelectronic devices. Furthermore, the gain mechanism relying on the combined action of the photogating and NC effects can be further exploited for the ultrasensitive few-photon detection technique and generally applicable to a large amount of low-dimensional semiconductors.

Methods

Fabrication of the multilayer MoS2 phototransistor

A 40 nm-thick TiN layer was deposited by ion beam sputtering on 300 nm-thick SiO2 grown on p-doped silicon substrates. The 10 nm-thick HZO thin film was grown on the TiN gate electrode by ALD process at 280 °C substrate temperature, and crystalized with a 60 s anneal at 500 °C in a nitrogen atmosphere. Hf[N(C2H5)CH3]4, Zr[N(C2H5)CH3]4 and H2O were used as Hf precursor, Zr precursor and oxygen source, respectively. The Hf/Zr ratio was controlled by alternate deposition of one cycle HfO2 and one cycle ZrO2, which was confirmed by X-ray photoelectron spectroscopy (XPS). Then a 6 nm-thick Al2O3 layer was deposited on the HZO film by electron beam evaporation. Multilayer MoS2 nano-sheets were mechanically exfoliated from bulk MoS2 crystals and mounted on the Al2O3 layer. Finally, the source (S) and drain (D) metal electrodes were fabricated using electron-beam lithography, thermal evaporation of Cr/Au films (15 nm / 45 nm) and lift-off.

Polarization of the ferroelectric HZO Film

In MoS2 NC phototransistors, the source electrode is connected to ground, and the polarization state of the ferroelectric HZO film in the gate stake is controlled by the gate voltage Vg, specific as follows. For the electrical measurement, taking Fig. 2e as an example, during the IdsVg measurement (IgVg is measured simultaneously and the results are shown in Supplementary Fig. 9), the source electrode is connected to ground, the drain voltage is constant (Vds = 0.1 V), and the gate voltage sweeps forward (Vg is from –1.9 to 1.9 V) and then reverse (from 1.9 to −1.9 V). As Vg sweeps forward, the ferroelectric polarization switches from downward to upward, the channel electrons are gradually accumulated and the device is turned on. As Vg sweeps reverse, the ferroelectric polarization switches from upward to downward, the channel electrons are gradually depleted and the device is turned off. For the optoelectronic measurement, taking Fig. 4a as an example, the source electrode is connected to ground, the operating voltages Vds and Vg are set as constant, and specially the gate voltage equals to the threshold voltage of the dark state (Vg = VT,dark = −1.6 V), yielding that the ferroelectric layer is polarized downwards and the dark current is suppressed to an extremely low level.

Measurement of electrical and optoelectronic properties

The electrical and optoelectronic measurement were carried out with a Lake Shore probe station and a Keysight B1500A semiconductor device analyzer under ambient conditions at room temperature. The photoresponse was tested under illumination of 520 nm/637 nm monochromatic light sources in a dark room. The light source was generated, switched, and transmitted by a Thorlabs ITC4001 laser diode/temperature controller and a LM9LP fiber-pigtailed laser diode. The spot diameter of the light source on the device surface was 1 mm, and the multilayer MoS2 nano-sheet of the device was positioned at the center of the spot. A high-speed Tektronix MDO3014 mixed domain oscilloscope was used to measure the time-resolved photoresponse.