Metallic-Phase-Fe$_3$GaTe$_2$ Enabled Interface Engineering for Self-Powered and High-Gain WS$_2$ Photodetectors

Introduction

Two-dimensional transition-metal dichalcogenides (TMDCs) have emerged as promising materials for next-generation photodetectors because of their strong light–matter interaction, atomically thin geometry, and compatibility with van der Waals (vdW) integration [1, 2, 3, 4, 5, 6, 7, 8, 9, ?]. Among them, WS$_2$ is particularly attractive owing to its visible-range optical absorption, thickness-dependent electronic structure, and stable photoresponse at room temperature [?, ?]. However, the practical performance of WS$_2$-based photodetectors is still strongly affected by inefficient photocarrier separation, interfacial recombination, and contact-related losses, which often require an external bias to obtain high photoresponse [?, ?]. Interface engineering provides an effective route to overcome these limitations. In vdW heterostructures (HS), band alignment and interfacial charge transfer can be tailored without the constraints of lattice matching, enabling the formation of built-in electric fields for efficient photocarrier separation and self-powered photodetection. At the same time, low-resistance contacts are essential for achieving high responsivity and detectivity.

Several contact-engineering strategies have been explored to improve carrier transport in two-dimensional optoelectronic devices. Graphene and black phosphorus electrodes can improve interfacial carrier transport, but they are often limited by high dark current and environmental instability, respectively [?, ?]. Conventional metal contacts such as Au, Ag, and Ti can form Schottky barriers at WS$_2$ interfaces, which impede carrier injection, increase contact resistance, and suppress photocurrent generation [?, ?, ?]. Low-work-function metals such as Sc and Ti can reduce barrier heights in TMDC devices, but oxidation and interfacial diffusion may compromise long-term stability [?, ?]. Metallic layered materials such as NbSe$_2$ and TaS$_2$ provide cleaner vdW interfaces, although nonideal work-function matching can still limit band alignment and carrier extraction [?, ?]. In this context, Fe$_3$GaTe$_2$ (FGT), a metallic vdW ferromagnet with high electrical conductivity and a favorable work-function relationship with WS$_2$, offers a promising interfacial contact for efficient charge separation and extraction [?, ?]. Compared with conventional evaporated metals and graphene electrodes, FGT can form a cleaner WS$_2$ interface, potentially reducing contact-related losses and supporting built-in-field-assisted photodetection [?, ?]. However, FGT-enabled interface engineering for self-powered and high-gain WS$_2$ photodetection remains largely unexplored.

In this work, we demonstrate a self-powered photodetector based on a WS$_2$/FGT vdW metal–semiconductor HS that enables efficient interfacial charge separation and zero-bias photoresponse. Owing to favorable band alignment at the WS$_2$/FGT interface and asymmetric contact-induced internal electric fields, the device exhibits a high self-powered responsivity of 23.5 A/W under 450 nm illumination. The apparent external quantum efficiency exceeding 100% is consistent with an internal gain mechanism, likely involving trap-assisted photogating in multilayer WS$_2$. The photodetector further maintains zero-bias visible-wavelength operation, with responsivities of 19.8 A/W at 520 nm and 6.1 A/W at 633 nm. These results demonstrate that WS$_2$/FGT vdW HS provide an effective platform for realizing low-power and high-sensitivity photodetection based on engineered metal–semiconductor interfaces.

Results and Discussion

Figure 1 schematically illustrates the device configuration, in which multilayer WS$_2$ partially overlaps with metallic FGT while the two materials are connected to asymmetric Au electrodes. The optical microscopy image in Figure 1 confirms the spatial overlap between WS$_2$ and FGT, while the corresponding atomic force microscopy (AFM) analysis in Figure 1 reveals well-defined flake thickness and a clean interface. Detailed structural characterization of the individual FGT and WS$_2$ crystals, together with the HS height profile, is provided in Supporting Information Figures S1 and S2.

Raman spectroscopy observation in Figure 1 illustrates characteristic vibrational modes corresponding to both FGT and WS$_2$. The in-plane ($\mathrm{E}^{1}_{2g}$, $\mathrm{E}^{2}_{2g}$) and out-plane ($\mathrm{A}^{1}_{2g}$) FGT modes at 101, 122.8, and 141.5 cm$^{-1}$, respectively are consistent with previously reported observation [?]. Similarly, the well-known WS$_2$ ($\mathrm{E}_{2g}$ =351.2 cm$^{-1}$) and ($\mathrm{A}_{1g}$ =422.1 cm$^{-1}$) modes shows a slight shift after integration with FGT, indicating interfacial interaction between the two layers that may arise from charge transfer and vdW coupling effects [?, ?].

The integrated intensity maps (Figure 1e and f) reveal spatial distribution of the two materials and clearly identify the overlapped HS region. The localized reduction in WS$_2$ Raman intensity within the HS overlapped region indicates that the metallic FGT layer alters the local optical and electronic environment of the overlying WS$_2$ layer. This attenuation can be associated with interfacial charge transfer and enhanced dielectric/electrostatic screening from the metallic FGT layer [?]. Consistently, the photoluminescence (PL) intensity of WS$_2$ is also quenched after contact with FGT, as shown in Supporting Information Figure S3, suggesting enhanced nonradiative relaxation and efficient interfacial carrier transfer. These spectroscopic results confirm the formation of a coupled WS$_2$/FGT vdW interface, which provides the structural basis for photodetection discussed below.

Photoresponse of the WS$_2$/FGT HS was systematically investigated by current–voltage ($I$–$V$) measurements under controlled illumination at room temperature, as shown in Figure 2. The drain–source current ($I_{\mathrm{ds}}$) was recorded as a function of drain–source voltage ($V_{\mathrm{ds}}$) from $-1$ to $+1$ V under dark conditions and under 450, 520, and 633 nm laser illumination, with grounded gate ($V_{\mathrm{g}} = 0$ V). As shown in fig:2a–fig:2c, the device exhibits pronounced asymmetric transport in the dark, with a rectification factor of 285.39 at $\lvert V_{\mathrm{ds}} \rvert= 1$ V, indicating the formation of an asymmetric Schottky-type junction [?]. This rectifying behavior arises from the $\mathrm{WS}_2/\mathrm{FGT}$ metal–semiconductor interface together with the asymmetric Au/$\mathrm{WS}_2$/FGT contact geometry. The lower work function of FGT relative to $\mathrm{WS}_2$ favors interfacial charge redistribution and band bending, generating a built-in electric field that assists photocarrier separation. Under illumination, the current increases significantly for all excitation wavelengths, confirming efficient photogeneration and carrier extraction in the HS.

The photoresponsivity “R” of the device, defined as the electrical output per unit incident optical power, was calculated using the relation [?].

Here $I_{\mathrm{ph}} = I_{\mathrm{light}} - I_{\mathrm{dark}}$ is the net photocurrent, $P_{\mathrm{light}}$ denotes the incident optical power density, and $A=80\ \mu\mathrm{m}^{2}$ is the effective photosensitive area of the $\mathrm{WS}_2$/FGT heterostructure. At $-1$ V, the device achieves high responsivities of $9.7 \times10^{3}$, $6.5 \times10^{3}$, and $2.5 \times10^{3}$ A/W under 450, 520, and 633 nm illumination, respectively (fig:2d). The stronger response at shorter wavelength is consistent with enhanced optical absorption and more efficient photocarrier generation in multilayer $\mathrm{WS}_2$. For all wavelengths, R decreases with increasing incident power, reflecting trap-state saturation and enhanced recombination at higher carrier densities. This sublinear behavior (also in Figure S4) is consistent with trap filling and suggests a contribution from photogating to the enhanced photoresponse [?, ?].

S. No.	Wavelength (nm)	Responsivity ($\mathrm{A}\mathrm{W}^{-1}$)	EQE (%)
Device 01	450	23.5	$6.4 \times10^{3}$
Device 01	520	19.8	$4.7 \times10^{3}$
Device 01	633	6.1	$1.2 \times10^{3}$
Device 02	450	20.2	$5.5 \times10^{3}$
Device 02	520	18.2	$4.3 \times10^{3}$
Device 02	633	5.3	$1.0 \times10^{3}$
Device 03	450	17.3	$4.7 \times10^{3}$
Device 03	520	14.2	$3.3 \times10^{3}$
Device 03	633	4.8	$0.9 \times10^{3}$

External quantum efficiency (EQE) was calculated as [?]

Where h, c, q, and $\lambda$ are Planck’s constant, the speed of light, the elementary charge, and the excitation wavelength, respectively. As shown in fig:2e, the apparent EQE follows the same power-dependent trend as R and reaches a maximum value of $2.6 \times10^{6}$ % under 450 nm illumination at $V_{\mathrm{ds}} = -1$ V. Although an apparent EQE exceeding 100% suggests an internal amplification process, EQE and photoconductive gain are conceptually distinct, and EQE alone does not provide a direct quantitative measurement of gain. Together with the sublinear power dependence, this result suggests a possible contribution from trap-assisted carrier dynamics in multilayer $\mathrm{WS}_2$.[?]

The noise-derived specific detectivity was calculated using the experimentally measured current-noise amplitude spectral density.[?]

Where $R$ is the responsivity, $A$ is the effective photosensitive area, and $I_n(f)$ is the measured current-noise amplitude spectral density. At $V_{\mathrm{ds}}=-1$ V and 100 Hz, $I_n=3.75\times10^{-13}$ A Hz$^{-1/2}$. Using $A=80$ $\mu\mathrm{m}^2$, the maximum noise derived specific detectivities are $2.3\times10^{13}$, $1.5\times10^{13}$, $5.9\times10^{12}$ Jones under 450, 520, and 633 nm illumination, respectively. Further details are provided in Supporting Information Figure S6. Motivated by the illumination-induced shift of the current-null point observed in fig:2a, fig:2b–fig:2c, we further evaluated the device performance under zero-bias conditions. As shown in Figure 3, the device maintains a pronounced power-density-dependent photoresponse at $V_{\mathrm{ds}}=0$ V, confirming that photocurrent generation does not require an externally applied voltage. The maximum zero-bias responsivities reach 23.5, 19.8, and 6.1 A/W under 450, 520, and 633 nm illumination, respectively (fig:3a). The corresponding apparent EQE values are approximately $6.4\times10^3$%, $4.7\times10^3$% and $1.2\times10^3$%, respectively (fig:3b). Both responsivity and EQE decrease with increasing incident power density, consistent with progressive trap filling and enhanced carrier recombination at higher photocarrier densities. Although these observations are consistent with internal-field-assisted self-powered operation, possible photothermoelectric and trap-assisted contributions cannot be completely excluded.

To understand the carrier dynamics and gain mechanism, incident-power dependence of the photocurrent was analyzed using the power-law relation $I_{\mathrm{ph}}\propto P^\alpha$. As shown in fig:4a, the photocurrent exhibits sublinear behavior at $V_{\mathrm{ds} }=-1$ V, with extracted $\alpha$ values of 0.60, 0.77, and 0.89 for 450, 520, and 633 nm excitation, respectively. The deviation from linearity is consistent with trap-assisted carrier dynamics and recombination processes in multilayer WS$_2$. The stronger sublinearity at 450 nm is consistent with more efficient photocarrier generation, faster trap filling, and partial screening of the built-in field at higher carrier density. In contrast, the larger $\alpha$ value at 633 nm indicates weaker trap saturation because of the lower photogeneration rate. The operational stability of the device was evaluated under periodically modulated illumination, as shown in fig:4b. The device exhibits stable and reproducible switching behavior for all three wavelengths, with on/off ratios of approximately $1.1 \times10^{4}$, $7.9 \times10^{3}$, and $2.8 \times10^{3}$ for 450, 520, and 633 nm, respectively. This wavelength-dependent trend agrees with the R and D results, confirming that the photoresponse is governed mainly by optical absorption in multilayer WS$_2$ and efficient carrier separation at the WS$_2$/FGT interface.

The transient response was measured under 450 nm illumination at $V_{\mathrm{ds}} = -1$ V. As shown in fig:4c, the device shows a clear and repeatable photocurrent response, with rise and fall times of 30 and 50 ms, respectively, extracted between 10% and 90% of the steady-state photocurrent. These response times are consistent with efficient carrier separation and extraction through the HS, while the remaining temporal response may involve trap-mediated carrier capture and release in multilayer $\mathrm{WS}_2$. This interpretation is consistent with sublinear power dependence and supports a possible trap-assisted contribution to the photoresponse. The performance of the $\mathrm{WS}_2$/FGT HS was benchmarked against representative 2D material-based photodetectors, as summarized in fig:4d and Table S2. Under zero-bias operation, the present device reaches a responsivity of 23.5 A/W under 450 nm illumination, placing it among competitive self-powered 2D photodetectors.

Material	Operation Condition ($V_\mathrm{g}$, $V_\mathrm{ds}$)	Wavelength (nm)	Responsivity (A W$^{-1}$)	Ref.
CdSe/Graphene	0	633	$10 \times10^{-3}$	[1]
Si/Graphene	0	850	$730 \times10^{-3}$	[2]
MoS$_2$/WS$_2$	0	532	$4.36 \times10^{-3}$	[3]
Graphene/MoSe$_2$/Au	0	450	$89.5 \times10^{-3}$	[4]
BiVO$_4$/MXene	0	447	$790.2 \times10^{-3}$	[5]
GaTe-MoS$_2$	0	514	1.36	[6]
MoS$_2$/BP	0	532	22.3	[7]
GeAs/InSe	0	405	0.36	[8]
MoSe$_2$/FePS$_3$	0	522	$52 \times10^{-3}$	[9]
WS$_2$/FGT	0	450	23.5	This Work
WS$_2$/FGT	0	520	19.8	This Work
WS$_2$/FGT	0	633	6	This Work

To elucidate the interfacial origin of the self-powered and high gain photoresponse, density functional theory (DFT) calculations and kelvin probe force microscopy (KPFM) measurements were used to examine the electronic structure and band alignment of the $\mathrm{WS}_2$/FGT HS. fig:5a shows the calculated band structure and density of states of multilayer $\mathrm{WS}_2$, confirming its indirect semiconducting nature with a bandgap of approximately 1.35 eV. The valence-band maximum is located near the $\Gamma$ point, while the conduction-band minimum lies near the K point, consistent with the visible-range optical absorption of multilayer $\mathrm{WS}_2$ [?, ?]. In contrast, FGT exhibits metallic electronic behavior, with spin-polarized bands crossing the Fermi level (fig:5b). The corresponding density of states shows a clear asymmetry between spin-up and spin-down channels, confirming the ferromagnetic metallic character of FGT [?, ?]. The projected density of states indicates that the electronic and magnetic properties of FGT are mainly governed by Fe 3d orbitals, as shown in Supporting Information Figure S8. This metallicity is essential for device operation because FGT acts as an efficient vdW carrier-collecting electrode at the $\mathrm{WS}_2$ interface.

The interfacial energy alignment was evaluated using KPFM measurements and DFT-calculated work functions. Experimentally, KPFM gives work functions of 4.557 eV for $\mathrm{WS}_2$ and 4.497 eV for FGT, corresponding to a relative offset of approximately 0.06 eV, as summarized in Supporting Information Figure S5. DFT calculations show the same relative trend (Figure S8), with $\mathrm{WS}_2$ having a higher work function than FGT, although the absolute values differ because of surface adsorbates, defects, interface dipoles, and idealized calculation conditions. Based on these results, the band diagram in fig:5c illustrates the formation of an asymmetric $\mathrm{WS}_2$/FGT metal–semiconductor junction. After contact, electrons transfer from lower-work-function FGT toward $\mathrm{WS}_2$ until Fermi-level equilibration is reached, producing band bending in $\mathrm{WS}_2$ and establishing a built-in electric field across the interface [?, ?, ?]. Under illumination, this field separates photogenerated electron–hole pairs and enables carrier collection even at zero external bias [?, ?].

Although the KPFM-derived work-function difference (~60 meV) is more than twice the thermal energy at room temperature ($k_{\mathrm{B}}T \approx26$ meV), this comparison does not determine the magnitude of the internal electric field. Therefore, the high zero-bias photoresponse is attributed to the combined effects of interfacial band bending, possible interface dipoles, asymmetric Au/WS$_2$/FGT contact geometry, efficient carrier extraction through metallic FGT, and a possible contribution from photogating-assisted carrier-lifetime extension in multilayer WS$_2$ [?, ?]. Together, the DFT and KPFM results support the relative work-function alignment and direction of interfacial charge redistribution at the WS$_2$/FGT interface. Combined with the rectifying transport characteristics and finite zero-bias photocurrent, these results are consistent with an internal electric field that assists photocarrier separation. Metallic FGT facilitates carrier extraction, while trap states in multilayer WS$_2$ provide photogating-assisted gain. These coupled effects explain the high responsivity, large EQE, and zero-bias photoresponse of the WS$_2$/FGT photodetector.

Conclusions

In summary, we demonstrate a self-powered WS$_2$/FGT vdW metal–semiconductor photodetector with high responsivity and large apparent EQE under weak visible illumination. Under 450 nm illumination, the device achieves a zero-bias responsivity of 23.5 A/W and an apparent EQE of $6.4 \times10^{3}\%$. At $V_{\mathrm{ds}} = -1$ V, it reaches a responsivity of $9.7 \times10^{3}$ A/W and a noise-derived specific detectivity of $2.3 \times10^{13}$ Jones at 100 Hz. The enhanced performance is attributed to interfacial carrier separation and efficient carrier extraction through metallic FGT, together with a likely contribution from trap-assisted photogating in multilayer WS$_2$. KPFM and DFT analyses further support favorable relative work-function alignment and interfacial charge redistribution between FGT and WS$_2$. These results establish FGT-enabled interface engineering as an effective strategy for realizing high-sensitivity, low-power photodetectors based on two-dimensional van der Waals heterostructures.

Associated Content

Supporting Information. Additional structural characterization of FGT and WS$_2$ crystals; AFM, SEM, and EDX analysis of the WS$_2$/FGT heterostructure; PL characterization; device reproducibility; KPFM work-function analysis; noise spectral density measurements; additional DFT details; experimental methods; and theoretical calculation details.

Author Information

Corresponding Authors

Ziwei Li – Hunan Institute of Optoelectronic Integration, College of Materials Science and Engineering, Hunan University, Changsha, P. R. China.

Email: [email protected]

Maciej R. Molas – University of Warsaw, Faculty of Physics, 02-093 Warsaw, Poland

Email: [email protected]

Authors

Wajid Ali – University of Warsaw, Faculty of Physics, 02-093 Warsaw, Poland.

Ming Huang– Hunan Institute of Optoelectronic Integration, College of Materials Science and Engineering, Hunan University, Changsha, P. R. China.

Juan Li– Hunan Institute of Optoelectronic Integration, College of Materials Science and Engineering, Hunan University, Changsha, P. R. China.

Jianhua Huang– Hunan Institute of Optoelectronic Integration, College of Materials Science and Engineering, Hunan University, Changsha, P. R. China.

Liuli Yang– Hunan Institute of Optoelectronic Integration, College of Materials Science and Engineering, Hunan University, Changsha, P. R. China.

Sajid Ur Rehman–School of Science, Minzu University of China, Beijing, P. R. China

Chinmay K. Mohanty– University of Warsaw, Faculty of Physics, 02-093 Warsaw, Poland.

Zahir Muhammad – National Key Laboratory of Spintronics, Hangzhou International Innovation Institute, Beihang University, Hangzhou 311115, P.R. China.

Author contributions

Z.L. and M.R.M. planned and supervised the project. W.A. and J.L. conducted experiments regarding the material synthesis, and characterization of the devices. M.H. and L.Y. supported to building optical setup for Raman and PL measurements. W.A. and C.M. contributed to data analysis and figure preparation. S.U.R. and Z.M. conducted the measurement of thermomagnetic curve and DFT calculations, respectively. W.A., Z.L. and M.R.M. wrote the manuscript with input from all co-authors.

† These authors contributed equally.

Competing Interest

The authors declare no competing financial interests.

Funding

This work is supported by the European Union’s Marie Skłodowska-Curie Actions under the Horizon Europe research and innovation programme (Grant No. 101204286). M.R.Molas gratefully acknowledges support from the National Science Centre, Poland (Grant No. 2022/46/E/ST3/00166 and 2023/50/O/ST3/00310). The work is also supported by the National Key Research and Development Program of China (Grant No. 2022YFA1203801), the National Natural Science Foundation of China (Grant No. 62275076), and the Hunan Provincial Science and Technology Department (Grant Nos. 2026JJ60487, 2023RC3093).

Supporting Information

Growth and Characterization of FGT

High-purity, gallium (Ga) foil (99.99%, Sigma-Aldrich), iron (Fe) powder (99.98%), and tellurium (Te) powder (99.98%, Alfa Aesar) were used as precursors for $\mathrm{Fe}_3\mathrm{Ga}\mathrm{Te}_2$ crystal growth. FGT single crystals were synthesized using a self-flux method. High-purity Fe, Ga, and Te were mixed in a 1:1:2 ratio, sealed in an evacuated quartz tube, heated to 1000 °C for 3 hours, held for 24 hours, and then slowly cooled to 770 °C over 100 hours. fig:S1a shows the X-ray diffraction (XRD) pattern of the FGT bulk crystal, oriented along the (00$l$) direction. This orientation indicates that the crystal's surface is perpendicular to the $c$-axis, aligning the plate-like surface with the $ab$ plane. The distinct and intense diffraction peaks signify the high quality of the directly grown samples. Advanced lattice parameter refinement and crystal structure elucidation were achieved through single crystal XRD analysis. fig:S1b illustrates the energy-dispersive X-ray (EDX) spectrum, which was utilized to ascertain the elemental ratios within the crystal. The elemental composition was deduced from the integrated intensity ratios of the distinct peaks corresponding to each element. The average of these peak intensities across three different types of peaks provided a composition estimate that closely aligns with the target composition for FGT. fig:1c and fig:1d shows the isothermal magnetization $M(H)$ curves measured at 300 K and 200 K, respectively. These curves clearly exhibit the characteristic ferromagnetic ordering behavior of FGT, as evidenced by the well-defined hysteresis loops. The data confirm the robust ferromagnetic nature of FGT at these temperatures, consistent with its known magnetic properties. Furthermore, these results are in excellent agreement with previously reported studies, reinforcing the reliability of the measurements and the intrinsic ferromagnetic behavior of FGT.

AFM analysis of WS$_2$/FGT

Surface morphology and thickness of the WS$_2$/FGT heterostructure were characterized using a Bruker Dimension Icon atomic force microscope operating in tapping mode. The corresponding height profile of WS$_2$ and FGT are shown in Figure S2(a, b).

Photoluminescence of $WS_2$/FGT heterojunction

The $WS_2$/FGT heterostructure devices were fabricated using mechanical exfoliation and dry-transfer technique. Multilayer $WS_2$ and FGT flakes were exfoliated using adhesive tape and first transferred onto polydimethylsiloxane (PDMS) substrates. Suitable flakes were identified using optical microscopy prior to transfer. The FGT flake was first transferred onto a pre-patterned Au electrode fabricated by standard photolithography followed by electron-beam evaporation. Subsequently, a multilayer $WS_2$ flake was transferred such that it partially overlapped the FGT flake while extending onto a second Au electrode, forming a $WS_2$/FGT metal–semiconductor HS device.

PL, Raman spectroscopy, and spatial mapping measurements were performed using a WITec alpha-300 confocal microscope equipped with a UHTS300 spectrometer. A 532 nm excitation laser was used together with a 50× objective lens (Zeiss, NA = 0.75), providing an effective laser spot size of approximately 1 $\mu$m. The laser power was maintained at 0.25 mW during measurements. An 1800 grooves mm$^{-1}$ grating was used for Raman spectral acquisition. The noise current spectra were measured using a KEYSIGHT 35670A dynamic signal analyzer. To reduce external electromagnetic interference, the device was enclosed inside a shielded metal measurement chamber during noise characterization.

Comparison of I–V curves of various $WS_2$/FGT devices

Electrical transport measurements of the $WS_2$/FGT heterostructure devices were performed using a CRX-VF probe station equipped with an Agilent B1500 semiconductor parameter analyzer. Prior to measurement, the device chamber was evacuated to reduce the influence of ambient adsorbates on electrical performance. A two-probe configuration was employed, in which the drain–source voltage ($V_{\mathrm{ds}}$) was applied across the $WS_2$/FGT HS device, and the corresponding current ($I_{\mathrm{ds}}$) was recorded. The voltage sweep range was from $-1$ V to $+1$ V to evaluate the current–voltage characteristics of the device. No external gate voltage was applied during measurements ($V_{\mathrm{g}} = 0$ V), allowing characterization of the intrinsic transport behavior of the $\mathrm{WS}_2$/FGT HS. All measurements were performed at room temperature.

KPFM image of the WS$_2$/FGT heterostructure

The surface potential distribution across the heterostructure was investigated using Kelvin probe force microscopy (KPFM) in AC mode with a conductive probe. The contact potential difference (CPD) between the tip and sample surface was measured to evaluate the relative work-function variation across the WS$_2$/FGT interface. To calibrate the tip, a gold film was used as a reference material, giving an electronic surface potential difference of gold ($\mathrm{eSPD}_{\mathrm{Au}}$) = − 0.07793 V = − 77.93 mV. The tip work function $\Phi_{\mathrm{tip}}$ was calculated as follows:

fig:S5a and S5b shows the KPFM image of the WS$_2$/FGT heterostructure and the corresponding surface potential profile. From this we determined the electronic surface potential differences to be $\mathrm{eSPD}_{\mathrm{WS}_{2}} = 0.465$ V and $\mathrm{eSPD}_{\mathrm{FGT}} = 0.525$ V. The work functions for WS$_2$ and FGT were then calculated as:

The lower work function of FGT relative to $WS_2$ indicates that electrons tend to transfer from FGT to $WS_2$ until Fermi-level equilibrium is reached, resulting in downward band bending on the $WS_2$ side near the FGT interface and the formation of an interfacial built-in electric field.

Measured noise spectral density of the $WS_2$/FGT device

The specific detectivity $D^{*}$ is a crucial figure of merit that quantifies the sensitivity of a photodetector, reflecting its ability to discern weak optical signals from inherent noise. This parameter allows for direct comparisons of device performance across varying active areas and geometries, providing insights into the minimum detectable signal. In our device, the measured current noise spectral density exhibits a characteristic $1/f$ dependence at low frequencies, indicating dominant flicker-noise contributions (Figure S6). At a modulation frequency of 100 Hz, the experimental noise current density $I_n$ is 3.75$\times10^{-13}$ A Hz$^{-1/2}$. Based on this measured noise, the Noise Equivalent Power (NEP) can be calculated as:

where $R = 9.7 \times10^3$ A/W is the responsivity of the device at $V_{ds} = -1$ V. Substituting these values:

The specific detectivity $D^*$ is then expressed as:

Here, $A = 80\ \mu\mathrm{m}^2$ represents the effective area of the photodetector, and $f_b$ denotes the noise frequency bandwidth (1 Hz). Substituting the experimental NEP yields:

Comparison of photoresponse of the WS$_2$/FGT device with the counterpart devices

DFT calculated work functions of WS$_2$ and FGT

The electronic structures of pristine WS$_2$, FGT, and the WS$_2$/FGT heterostructure were calculated using density functional theory (DFT) as implemented in the Vienna Ab initio Simulation Package (VASP). Spin-polarized calculations were performed to account for the intrinsic ferromagnetic nature of FGT. A vacuum layer of $20\,\text{\AA}$ was introduced along the out-of-plane direction to eliminate artificial interactions between periodic slabs. The interlayer van der Waals interaction between WS$_2$ and FGT was described using the Grimme DFT-D3 correction method. The on-site Coulomb interaction of Fe $3d$ orbitals was treated using the DFT + U approach with an effective Hubbard parameter $U_{\mathrm{eff}} = 4$ eV. The exchange–correlation interaction was described using the Perdew–Burke–Ernzerhof (PBE) functional within the generalized gradient approximation (GGA), and the plane-wave cutoff energy was set to 500 eV. All atomic structures were fully relaxed until the residual forces were less than $10^{-2}$ eV $\text{\AA}^{-1}$ and the total energy convergence criterion reached $10^{-6}$ eV. The Brillouin zone was sampled using a $\Gamma$–centered $11 \times11 \times1$ Monkhorst–Pack k–point mesh.