In recent years，with the breakthrough of emerging technologies such as photonic integrated circuits，photodetectors，as a component of many optical and optoelectronic devices，have made great progress.^［1-3］ These technological advances have not only enhanced the performance of photodetectors but also expanded their application scope. They continually drive the development of photodetectors towards faster response speeds，lower dark currents，and broader spectral responses. Nevertheless，conventional photodetectors utilizing bulk substrates such as silicon^［4］，germanium^［5］，and III-V group semiconductors^［6］ encounter challenges in concurrently achieving a swift photonic response，minimizing dark current，and ensuring a broad spectral sensitivity. An emerging and viable alternative is found in the realm of two-dimensional（2D） materials. These materials are characterized by their robust interaction with light，adjustable energy bandgaps，and harmonization with current semiconductor manufacturing processes，alongside the considerable advantages presented by their atomically precise heterointerfaces^［7-8］. For instance，2D material-based photodetectors with ultra-high response speeds（up to 500 GHz） have been well-documented，achieving high-speed responses through the exceptionally high mobility of graphene^［9］. However，these devices also exhibit milliampere-level dark currents. To reduce the dark current and enhance the response speed of these devices，researchers have employed p-n junctions^［3，10］. However，due to material growth and the extended length of the depletion zone，there is potential for further improvement in the response speed of the devices^［10-11］.

In this work，we report a photodetector fabricated using multilayer MoTe₂ to form a p-n homojunction and photovoltaic mechanism. By adopting a poly（vinylidene fluoride-trifluoroethylene）（P（VDF-TrFE）） ferroelectric field and a semi-floating gate structure composed of graphene and hexagonal boron nitride，we successfully reduced the junction length，achieving ultra-low dark current and high-speed response. The unique properties of 2D materials allowed us to modulate doping of electrons or holes simply by applying an electrostatic field. The floating gate part，made of graphene and hexagonal boron nitride，effectively shielded the MoTe₂ channel from the polarization effects of P（VDF-TrFE），efficiently forming an in-plane p-n junction. Unlike traditional rigid ferroelectric materials like LiNbO₃ and BiFeO₃，organic ferroelectric copolymers such as P（VDF-TrFE） exhibit exceptional flexibility，transparency，retention，and durability，offering new possibilities for the widespread application of 2D material-based p-n junctions^［12-14］. These materials can be easily prepared through low-temperature and simple processes like spin-coating，potentially reducing the cost and simplifying the manufacturing process for large-scale production of 2D material p-n junctions. Leveraging the depolarization field of P（VDF-TrFE）^［15］，we altered the bandgap of MoTe₂，achieving efficient photodetection at a wavelength of 1550 nm.

Fig. 1（a） presents a schematic of the proposed device structure. The layered composition，from bottom to top，consists of an aluminum（Al） gate electrode，a P（VDF-TrFE） dielectric layer，a partially covered graphene floating gate，and a h-BN tunneling layer. Such a configuration facilitates efficient charge storage and enables effective modulation of carrier concentration in the MoTe₂ channel. In this structure，MoTe₂ serves as the photosensitive channel layer，responsible for sensing optical signals. Analogous to the structure of MoS₂，MoTe₂ comprises Te-Mo-Te layers bonded by van der Waals forces^［16］. MoTe₂ is a two-dimensional semiconductor with a bandgap ranging from 0.93 to 1.1 eV as its thickness decreases from bulk to a single layer^［17］. The bipolar characteristics of MoTe₂ makes it easy to modulate to a p-type or n-type semiconductor by ferroelectric polarization. Other bipolar two-dimensional materials such as WSe₂ and BP were considered but they are vulnerable to ambient environment. It is noteworthy that the graphene and h-BN layers only cover half of the channel，indicating that the floating gate exerts control over only half of the channel region. This design feature allows for selective doping of the channel region through the ferroelectric polarization effect or gate voltage，thereby tuning its conductive properties.

In the device architecture，the P（VDF-TrFE） layer serves a dual purpose. Firstly，it provides a ferroelectric field which electrostatically dopes the MoTe₂，inducing selective doping due to the influence of the graphene and h-BN floating gate，resulting in the formation of a p-n junction. Secondly，the electric field generated by the P（VDF-TrFE） layer modifies the energy bandgap of the MoTe₂，thereby extending its spectral response. ^{［15，18］}The strategic placement and engineering of the P（VDF-TrFE） layer are critical to achieving the desired modulation of the device's electrical and optical properties. First，the two-dimensional materials were characterized by Raman spectroscopy，as shown in Fig. 1c. Individual MoTe₂ exhibited typical Raman peaks，indicating they are few-layer structures^［19］.The precise thicknesses were further studied through Atomic Force Microscopy measurements，revealing that the thickness of molybdenum ditelluride is 10nm，and the thickness of the floating gate is 5 nm. To gain a deeper understanding of the regulatory effect of the P（VDF-TrFE） dielectric layer on the electrical properties of two-dimensional MoTe₂，this study systematically analyzed the transfer characteristic curves of the channel under different coverage conditions.

In the transfer characteristic curve depicted in Fig. 2（a），the device is characterized by a channel that is not encapsulated by graphene and boron nitride，employing solely P（VDF-TrFE） as the dielectric layer in the Ferroelectric Field-Effect Transistor（FeFET）. Under a source-drain bias（$V_{\mathrm{s}\mathrm{d}}$） of ±0.1 V，the transfer characteristic curve distinctly exhibits the bipolar semiconductor properties of MoTe₂^［20］. The counterclockwise hysteresis observed in the current reflects the ferroelectric nature of P（VDF-TrFE）^［18］. Changes in the gate voltage alter the polarization direction of P（VDF-TrFE），and upon the removal of the gate bias，the residual polarization of P（VDF-TrFE） leads to the counterclockwise hysteresis in the current-voltage curves.In the experiments presented in Figs. 2b and 2c，we initially applied gate voltages of -20 V and +20 V to polarize P（VDF-TrFE）. Subsequently，we tested the back-gate transfer characteristics of the floating gate ferroelectric field-effect transistors（FGFeFET） and FeFET under various polarization states（$V_{\mathrm{s}\mathrm{d}}$ set at 0.5 V）. In Fig. 2b，the minimum points of the back-gate transfer curves for channels under different control methods were -3.3 V and 8.6 V，respectively. In Fig. 2c，these points were -4.3 V and 0.5 V，respectively. The concentration and type of carriers were calculated using the following formula^［21］ ： $n=q\times (V_{\mathrm{g}}-V_{\mathrm{n}\mathrm{g}})\times C_{\mathrm{g}}$. Here，$V_{\mathrm{g}}$ represents the applied back-gate voltage，which was set to zero（$V_{\mathrm{g}}$=0 V） in this experiment，and $V_{\mathrm{n}\mathrm{g}}$ is the gate voltage corresponding to the lowest channel conductance. $C_{\mathrm{g}}$ indicates the capacitance of silicon dioxide（SiO₂），and $q$ is the charge carried by an electron. The sign of n indicates whether the majority carriers are electrons or holes. The results show that under different polarization states of P（VDF-TrFE），channels not covered by graphene and hexagonal boron nitride exhibited different types of carriers compared to those that were covered.

The types of majority carriers in the MoTe₂ channels under different polarization states are shown in Fig. 2d. When the P（VDF-TrFE） is in the Pup state，the majority carriers in the FGFeFET are holes，with a concentration of 2.41×10¹⁷ cm^-3； in the FeFET field-effect transistor，the majority carriers are electrons，with a concentration of 5.78×10¹⁷ cm^-3. In the SFGFeFET，the depletion width of the p-n junction is 193 nm，with a barrier height of 0.74 eV. When the P（VDF-TrFE） is in the P_down state，the majority carriers in the FGFeFET are electrons，with a concentration of 6.06×10¹⁶ cm^-3； in the FeFET，the majority carriers are holes，with a concentration of 3.41×10¹⁷ cm^-3. In the SFGFeFET，the depletion width of the p-n junction is 167 nm，with a barrier height of 0.68 eV. The distinct carrier concentrations and types in the FeFET and FGFeFET are attributed to the change in graphene's Fermi level under the influence of the gate voltage and ferroelectric field，allowing carriers in the MoTe₂ channel to tunnel into the graphene. When the gate voltage is removed，the electric field generated by the graphene partially shields the effect of the P（VDF-TrFE） electric field. This results in different concentrations and types of majority carriers in the MoTe₂ channels covered and not covered by graphene，thus forming a p-n junction within the device.

Fig. 2（e） presents the transfer characteristics curves of the SFGFeFET under ±$V_{\mathrm{s}\mathrm{d}}$ conditions. A significant shift in the rectification direction of the device is observed when the gate voltage is switched from positive to negative. Additionally，the device exhibits a stable current response within the gate voltage ranges of -10 V to -20 V and 10 V to 20 V，indicating that the P（VDF-TrFE） ferroelectric layer possesses excellent data retention characteristics，and the floating gate structure provides a stable electric field environment. We also measured the output characteristics（$I_{\mathrm{s}\mathrm{d}}-V_{\mathrm{s}\mathrm{d}}$） of the SFGFeFET under various states，as shown in Fig. 2（f）. The output characteristics under different gate voltages exhibit similar rectifying behavior，with the magnitude and direction of rectification changing with the gate voltage. This shows that the direction of the built-in electric field of our device changes with the change of gate voltage. In conclusion，through the design of a P（VDF-TrFE） dielectric layer and a graphene/h-BN semi-floating gate，we can effectively modulate the MoTe₂ channel，achieving precise control of in-plane structures. This finding provides new strategies and theoretical foundations for the design and optimization of electronic devices based on two-dimensional materials

To verify the existence of the junction region within our device，we conducted photovoltaic current mapping at zero source-drain bias（$V_{\mathrm{s}\mathrm{d}}$=0 V）. The scanning photocurrent measurement principle diagram of SFGFeFET device is shown in Fig. 3（a）. The scan results are shown in Fig. 3（b）. Notably，a significant positive photocurrent is discernible at the boundaries of graphene，indicative of a depletion zone at these interfaces. The inherent electric field present facilitates the separation of photogenerated carriers，resulting in the generation of a spontaneous photocurrent. This effect is also observed near the right electrode，where the photocurrent emerges spontaneously and aligns with the electric field direction of the junction region. This phenomenon is attributed to a Schottky junction formed between the metal electrode and the molybdenum telluride，generating a current that must coincide with the direction of the p-n junction to yield spontaneous photocurrent. Furthermore，the optoelectronic performances were measured and depicted in Fig. 3（c）. The device exhibited robust responses across different light intensities. Additionally，we scrutinized the pure photocurrent response at 1 550 nm，as shown in Fig. 3（d）. Given that 1 550 nm does not fall within the intrinsic response spectrum of MoTe₂，these observations suggest that our device has successfully extended the responsive spectral range of molybdenum telluride. This extension is attributable to the Stark effect，where the depolarization field of of the P（VDF-TrFE） alters the bandgap of molybdenum telluride，^［30］ consequently broadening its spectral bandwidth^［31］. The Stark effect engenders a shift in the energy levels of the electronic states，thus enabling the absorption of photons with lower energies than the original bandgap，which is a remarkable demonstration of the tunable optoelectronic properties of our device. In Fig. 3（e），the photogenerated current signals exhibited rise and fall times of 2 μs each，with the limiting factor being attributed to our test equipment. However，with a device capacitance of only 1.79×10^-5 pF，the equivalent RC circuit response time is approximately 27 ps. Given a junction length of about 167 nm and an electric field strength of 4×10⁴ V·cm^-1，the carrier transit time is calculated to be around 3.7 ps. Consequently，this suggests a dynamic response bandwidth of approximately 5.8 GHz.

As revealed in Fig. 3（f），both the short-circuit current（$I_{\mathrm{s}\mathrm{c}}$） and open-circuit voltage（$V_{\mathrm{o}\mathrm{c}}$） exhibited an increasing trend with the enhancement of 520 nanometer light illumination power. When the device reached saturation，the $V_{\mathrm{o}\mathrm{c}}$ attained a maximum of 0.4 V，and the saturation current was measured at 71 nA. Fig. 3（g） details the power dependency of the photoresponsivity（$R$） and detectivity（$D^{*}$） under zero bias（$V_{\mathrm{s}\mathrm{d}}$=0 V）. The photoresponsivity is calculated using the formula $R=I_{\mathrm{p}\mathrm{h}}/p$，where $I_{\mathrm{p}\mathrm{h}}$_{represents the photogenerated current and $p$ denotes the power of illumination. Due to the depletion of carriers in the space charge region of the photodiode，thermal noise and generation-recombination（G-R） noise are negligible^［32］. The intrinsic noise is primarily contributed by shot noise in the junction current. Thus，the detectivity（$D^{*}$） of our photovoltaic detector is calculated using the formula $D^{*}=A^{\mathrm{1}/\mathrm{2}}R(\mathrm{2}q{I}_{\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{k}}{)}^{-\mathrm{1}/\mathrm{2}}$，^［33］ where A is the effective area of the photodetector（20×10^-8 cm²），q is the elementary charge，and $I_{\mathrm{d}\mathrm{a}\mathrm{r}\mathrm{k}}$​ is the dark current（0.1 pA）. At an illumination of 520 nm with a light power of 2.1 nW，the photodetector demonstrated a maximum photoresponsivity of 5 mA·W^-1 and a detectivity of 3.7×10¹⁰ Jones. To evaluate the performance of our SFGFeFET photodetector，Fig. 3（h） compares planar homojunction photodetectors made of different two-dimensional materials in terms of response time and dark current density. The ultra-low dark current density in our device may be attributed to the passivation of surface defects by P（VDF-TrFE） and boron nitride at the bottom，which suppresses the generation-recombination current，thereby reducing the dark current. The faster response speed of our device may be due to the short depletion region of the p-n junction due to the strong iron electric field in the device，which helps to reduce the carrier transfer time and thus improve the photoelectric response speed. The response speed，responsivity，and detectivity of the device can be further enhanced by selecting electrode materials that reduce contact resistance，or by extending the thickness MoTe2 to increase light-matter interaction.}

Fig. 4（a），we present the current characteristics of our device under various gate voltages（$V_{\mathrm{g}}$） and source-drain biases（$V_{\mathrm{s}\mathrm{d}}$），at a wavelength of 1 270 nm and under an illumination of 560 mW，contrasted with its dark state. The measurements reveal distinctive behavioral characteristics of the photocurrent and dark current in response to changes in $V_{\mathrm{g}}$ and $V_{\mathrm{s}\mathrm{d}}$，laying the experimental groundwork for the device's potential application in logical computation. In the configuration shown in Fig. 4（b），we utilized two structurally identical devices and selected an appropriate resistor R，based on the $V_{\mathrm{S}\mathrm{S}}$ magnitude，to maintain operation at a source-drain bias of -0.5 V and a gate voltage of 8 V. With $V_{\mathrm{S}\mathrm{S}}$ set at 5 V and $V_{\mathrm{D}\mathrm{D}}$ at 0 V，the ideal resistance was determined to be 1 GΩ，correlating with the current levels derived from the transfer characteristics. We have defined the illuminated state（filled circles） as a logical '1' and the dark state（unfilled circles） as a logical '0'，thus creating a binary input system based on the presence or absence of light. The equivalent circuit diagram and the computational results，as demonstrated in Fig. 4（c），indicate that the currents from both devices merge through a common resistor R，converting the current signal into a voltage output，culminating in a singular output signal. The results depicted in Fig. 4（c） confirm that if either device is illuminated，the output signal reaches a high level，effectively implementing an 'OR' logic operation.

Fig. 4（d） illustrates the circuitry devised to perform 'NAND' and 'XOR' logical operations. By connecting the source electrodes of the devices to positive and negative voltage sources，$V_{\mathrm{D}\mathrm{D}}$ and -$V_{\mathrm{D}\mathrm{D}}$，respectively，and assuming a total bus voltage of 10 V，we can set $V_{\mathrm{D}\mathrm{D}}$ to 10 V and -$V_{\mathrm{D}\mathrm{D}}$ to 0 V，with $V_{\mathrm{S}\mathrm{S}}$ at 5 V to achieve the desired bias. Appropriate resistor values ensure that both devices operate at source-drain voltages $V_{\mathrm{S}\mathrm{D}}$ of +0.5V with a gate voltage $V_{\mathrm{g}}$ of -10 V，and $V_{\mathrm{s}\mathrm{d}}$ of -0.5 V with $V_{\mathrm{g}}$ of +7 V，respectively. The equivalent circuit diagram and outcomes of these operations are presented in Fig. 4（e），where the devices under varying light input conditions yield distinct voltage outputs，thus realizing 'NAND' logic gate operations. By employing the same circuit design and setting the gate voltages of the two devices to -8 V and 5 V，we can switch the logic operation mode to an 'XOR' gate，as depicted in Fig. 4（f）. The flexibility of this design supports various logic computations，underscoring the potential of our device architecture for applications in photonic integrated circuits where light serves as the control parameter for executing logic tasks. The distinct voltage output levels under different light input conditions clearly delineate logical states，affirming the design's accuracy in performing complex logical operations.

The MoTe₂ p-n junction controlled by semi-floating ferroelectric gate is achieved. The diode exhibits robust rectification characteristics with a rectification ratio exceeding 10⁴. As a photodetector，our device demonstrates a high photoresponsivity of 5 mA·W^-1，a rapid response time of 2 µs，an open-circuit voltage of 0.4 eV，and a high specific detectivity of 3.7 × 10¹⁰ Jones for visible light（520 nm），all achieved without bias or gate voltage，implying low power consumption. Notably，at room temperature，the response spectrum extends to the short-wave infrared（SWIR） range（1 550 nm），thereby addressing the limitations of traditional Si-based and infrared photodetectors. Building upon our device architecture，we constructed a binary input system based on the presence or absence of illumination. These devices operate under specific voltage conditions，converting the current signals passing through them into voltage outputs to perform basic logic operations. Such a photonic-controlled logic computation system demonstrates significant potential in optoelectronic integrated circuit design. In this system，light signals are utilized not only for sensing information but also for directly participating in the logical decision-making process. The distinct voltage output levels under different light input states clearly delineate the logic states，affirming the high accuracy and reliability of this design in executing complex logic operations.