Polarization-sensitive photodetectors have attracted great interest because of their significant applications, such as in high-contrast polarizers, optical switches, polarization sensing, connector products, and optical radar, which have a wide range of application needs [1–7]. In early research, the polarization-sensitive photodetector mainly focused on the macroscopic anisotropy of the device, which requires complex molding and has a low anisotropy ratio, seriously restricting the practical application [8–10]. In addition, research has found that 2D materials have inherent anisotropy and have been promising for preparing polarization-sensitive photodetectors with high anisotropy ratio due to their unique photoelectric characteristics [11–16]. Graphene (Gr) is an outstanding representative of the 2D family, and prepared Gr-based photodetectors have excellent quantum efficiency and response time [17–19]. Unfortunately, due to the lack of anisotropic crystal structure, Gr does not exhibit intrinsic polarization sensitivity.

Depositing non-centrosymmetric metal structures on graphene is an effective strategy. For example, Yu et al. used graphene nanoribbons with periodic metal patterns to detect polarized light via localized plasmons [20]. Novoselov et al. achieved higher anisotropy by integrating finger-shaped plasmonic resonators [21]. It can be found that the reported anisotropy ratio of Gr-based polarization-sensitive photodetectors is relatively low and mainly concentrated in the visible to infrared band [22]. More importantly, there is a lack of systematic research on the deposition of non-centrosymmetric devices. Besides, there are few reports on polarization-sensitive graphene photodetectors in the ultraviolet region. Therefore, it is urgent to find a simple, stable, and poisonous strategy to solve the above problems.

In this work, $\mathrm{Gr}/{\mathrm{Al}}_2{\mathrm{O}}_3/\mathrm{GaN}$ tunnel junction photodetectors are realized by structuring heterojunctions, and different shapes and quantities of non-centrosymmetric metals are deposited on the Gr surface to achieve UV polarized light detection. Since the non-centrosymmetric metal structure breaks the original symmetric local field under the surface plasmon resonance and gives momentum to the photogenerated carrier under the high conductivity metal conduction, the device has the ability to detect polarized light. Therefore, it was discovered through systematic studies when a T-type metal structure is deposited, the device has an anisotropy ratio of 17.4 at 0 V and 21 at $-2\mathrm{V}$. While the E-type metal structure is deposited, the device has an anisotropy ratio of 1.9 at 0 V and 84.3 at $-2\mathrm{V}$, and with the increase in the number of non-centrosymmetric metal structures, the anisotropy ratio of the device also increases, which can be attributed to the photogenerated carrier gaining more momentum. Moreover, the effective repair of an ultra-shallow van der Waals heterojunction and interface engineering formed by Gr and GaN, the high photoelectric conversion efficiency of GaN, and the FNT effect together provide the possibility for the device to achieve a high anisotropy ratio. This exploratory study provides a valuable reference for developing and practically applying low-dimensional polarization-sensitive UV photodetectors.

The structure of the $\mathrm{Gr}/{\mathrm{Al}}_2{\mathrm{O}}_3/\mathrm{GaN}$ tunneling heterojunction UV polarization sensitive photodetector with a deposited E-type non-centrosymmetric metal structure is shown in Fig. 1(a), and Fig. 1(b) shows the top view of the device structure. The upper part of Fig. 1(a) shows how the incident polarized light is produced. The laser (325 nm) first passes through the polarizer and then through the reflector to uniformly reflect the polarized light on the surface of the device where the non-centrosymmetric metal pattern is deposited. During this process, the polarization state does not change, and the direction of vibration changes only with the rotation of the polarizer. Besides, in order to effectively repair the Gr/GaN interface and improve device performance, interface engineering techniques were used, and a smooth and flat ${\mathrm{Al}}_2{\mathrm{O}}_3$ passivation layer 1 nm thick was deposited on the GaN surface [Fig. 1(c)], which can efficiently reduce the interfacial defect. Electrodes of 200 nm were deposited on Gr and GaN surfaces respectively. And non-centrosymmetric metal structures (100 nm) with different morphologies and numbers were deposited on Gr. Meanwhile, two very small FWHW peaks located at 34.3° and 41.5° can be found from the XRD spectra of the ${\mathrm{Al}}_2{\mathrm{O}}_3/\mathrm{GaN}$ heterojunction in Fig. 1(d), corresponding to the wurtzite structure of the GaN (002) peak and Al (006) peak, indicating that both have high crystal quality [23]. The Gr Raman spectra show that the G/2D peak intensity ratio is about 1/2 [Fig. 1(e)]. This is consistent with the previous report [24], confirming that the transferred Gr is a better quality monolayer [25,26]. This lays a foundation for the preparation of high-performance devices. The corresponding band distribution diagram of the device is shown in Fig. 1(f). After introducing a thin ${\mathrm{Al}}_2{\mathrm{O}}_3$ layer between the Gr and GaN [Fig. 1(d)], the tunneling structure would be formed, and the energy-band diagram of the photodetector under reverse bias would be aligned as shown in Fig. 1(d). In this case, the dark current is expected to be suppressed due to the increased Schottky barrier height (SBH), and the photogenerated excess holes would accumulate at the insulator-semiconductor interface [27,28]. Further, when the insulation layer is thin enough, the impact ionization driven by high electric felid would subsequently occur to realize the photocurrent multiplication [28].

In order to systematically study the role of non-centrosymmetric metal structures in polarized light detection, the response characteristics of monolayer Gr, GaN, and Gr/GaN structures to ultraviolet polarized light were tested, as shown in Fig. 1(g). Unsurprisingly, the output photocurrent of the device remained almost unchanged by constantly changing the vibration direction of the polarized light. Moreover, the I-T curves of each structure at different polarization angles also show that the photocurrent does not change with the polarization angle, which further confirmed that the monolayer Gr and GaN films were not sensitive to UV-polarized light. Further, the polarization sensitivity of the E-type non-centrosymmetric metal structure $\mathrm{Gr}/{\mathrm{Al}}_2{\mathrm{O}}_3/\mathrm{GaN}$ tunneling heterojunction UV photodetector is studied. The vertical and horizontal lengths of the E-type metal are 7.5 μm and 8.5 μm, respectively, and the width and spacing of the three horizontal rods are 1.5 μm. The relationship between the output photocurrent of the device and the incident light power density under the irradiation of polarized light at different bias voltages and different polarization angles was studied, as shown in Fig. 1(h). The results show that the device has significant UV polarization sensitivity. When the polarized light angle is 0°, the output current of the device increases with the increase of the incident light power, while at 90°, the current hardly changes. In addition, as the bias increases, the anisotropy ratio between 0° and 90° polarization angles increases. At $-2\mathrm{V}$, it is close to two orders of magnitude. The results of polarized light sensitivity can be explained by the fact that on the one hand, under the guidance of the highly conductive non-centrosymmetric E-type metal structure on the Gr surface, the E-type metal electrode generates a large anisotropic electric field through the effect of the surface plasmon resonance. On the other hand, when the incident light irradiates on the Gr surface, it causes thermal absorption of light, resulting in the Seebeck effect. Therefore, under the Seebeck coefficient gradient and anisotropic electric field, the photogenerated carrier obtains momentum [29]. At this point, if the polarization angle is 0°, the polarized light will have a vertical vibration vector parallel to the direction of the built-in electric field so that the charge carriers generated by the light can obtain the maximum momentum, which is conducive to the separation of the charge carriers generated by the light. The direction of the bias applied at the same time coincides with the direction of the internal electric field, resulting in a very large photocurrent [30]. Furthermore, Fig. 1(i) depicts a remarkable anisotropic photo-response of the photodetector via the 3D colormap within the bias from 0 to $-2\mathrm{V}$. Obviously, as the polarization angle increases, the output photocurrent of the device continues to increase, reaching a maximum value at 180° (when the polarized light is incident along the $y$ direction). When the polarization angle continues to increase, the output photocurrent reaches the minimum at 270°. Good sinusoidal properties were demonstrated throughout the whole cycle, indicating that the prepared photodevices have excellent polarization sensitivity.

To further evaluate the polarization sensitivity of the device, the evolution of the current under zero bias was further extracted as a function of the polarization angle in polar coordinates, as plotted in Fig. 2(a). It is clear that the images show good symmetry and the output photocurrent changes periodically with the change of polarization angle. And a photocurrent anisotropy ratio of about 1.9 was achieved (defined as the ratio of the measured photocurrent at 90° polarization to 0°), which is similar to the results of other structures [31,32]. However, with the increase of applied bias, the anisotropy ratio of the device also increases significantly, reaching a maximum value of 84.3 at $-2\mathrm{V}$ [Fig. 2(b)], indicating that the prepared device is very sensitive to polarized light, and the image is becoming more and more symmetrical, as shown in Fig. 2(c). Under different bias voltages, the planar diagram of the relationship between the output photocurrent of the device and the polarization angle of the incident light has also been fitted and analyzed, as shown in Figs. 2(d)–2(f). All curves show standard sinusoids with a periodic change of decreasing first and then increasing, which is consistent with the previous analysis. Compared with the polar diagram, the planar graph shows a more obvious variation of the output photocurrent with the applied bias, indicating that the photogenerated carrier has higher separation efficiency and momentum under the synergistic action of the built-in electric field and the external bias, which makes the photocurrent larger. In order to further evaluate the stability and polarization sensitivity of the device, the real-time I-T curve of the output photocurrent with the polarization angle under different bias voltages was tested, as shown in Fig. 3. A very significant sinusoidal variation was observed, indicating that devices with a single E-type metal have good polarity dependence and stability. Besides, the influence of the number of E-type metals on the polarization sensitivity of the device is further studied (in order to ensure the accuracy and authenticity of the experiment, the size and spacing of all preparation processes and E-type metals are the same), and Fig. 4 shows the polarization-sensitive test of the device with two E-type metals. The output photocurrent curve has obvious symmetric and sinusoidal characteristics, indicating that the device has excellent polarization sensitivity.

Obviously, as shown in Fig. 5(a), with the increase of the number of metal E-types, the output photocurrent of the device also increases significantly, resulting in an increase in the anisotropy ratio of the device (single E-type metal structure: 84.3; two E-type metals structure: 115.5). This can be explained by the fact that as the range of anisotropic local electric fields increases, more polarized photothermal carriers are induced, resulting in an increase in photocurrent, indicating that the optical response of non-centrosymmetric metals is global. In addition, Fig. 5(b) shows the curve of the output photocurrent and time of the device depositing two E-type metals as the polarization angle of the incident light changes. The results show that the device has stable performance, good switching characteristics, and polarized light anisotropy.

Simulations [Fig. 6(a)] show that when 325 nm polarized light hits the E-type $\mathrm{metal}/\mathrm{Gr}/{\mathrm{Al}}_2{\mathrm{O}}_3$ structure, it triggers surface plasmon resonance. These oscillations generate localized hot spots with enhanced electric fields, which impart momentum to carriers along specific crystallographic axes [33,34]. Moreover, the range and intensity of electric field diffraction change significantly with the change of polarization angle. When the polarization angle is 0°, the plasmon propagates the widest range, and the maximum photogenerated current is obtained under the cooperative action of vertical electric field. The T-type metals (remove two horizontal rods of E-type metal) also exhibit a similar phenomenon, indicating the universality of non-centrosymmetric metal sensitivity to polarized light [Fig. 6(b)].

We further fabricated a T-type metal$/\mathrm{Gr}/{\mathrm{Al}}_2{\mathrm{O}}_3/\mathrm{GaN}$ junction polarized light detector. Similar to the E-type metal arrangement, the horizontal rods all face in the direction of the electrode (Fig. 7), which facilitates the photogenerated carrier to gain maximum momentum when the polarization angle of the polarized light is parallel to the direction of the built-in electric field.

The above results show that Gr/GaN photodetectors exhibit strong polarization sensitivity under the action of non-centrosymmetric metals. In order to further explain and understand the mechanism of non-centrosymmetric metal, the working principle of the device is analyzed in detail in this study. When Gr absorbs photons to generate hot carriers, the hot carriers will undergo different heat release processes during transmission and pass through the insulator barrier through the tunneling effect, which is absorbed by the electrode and contributes to the total current. In addition, when non-centrosymmetric metals are deposited on the surface of Gr, the metals will be doping Gr to a certain extent. Due to the local heating of light, the temperature distribution on the surface of Gr will be uneven, resulting in different Seebeck coefficients at different positions on the surface of Gr, forming a Seebeck coefficient gradient, causing the Seebeck effect, and generating thermal current [32]. Its direction and magnitude are determined by the resulting Seebeck coefficient gradient. The monolayer Gr is very thin and has a band gap of 0, so the Seebeck coefficient (S) generated in Gr can be approximately described by the Mott relation [38]: $$S=-\frac{{\pi }^2{K}_B^{2}T}{3e\sigma }\frac{\partial \sigma }{\partial \varphi },$$(1)where $K_B$, $T$, $\varphi$, $\sigma$, and $e$ are the Boltzmann constant, temperature, chemical potential, conductivity, and charge constant, respectively. It is obvious that S is proportional to $\sigma$. However, the electrical conductivity of Gr is related to the chemical potential. Through literature review, the expression is as follows [39]: $$\sigma =\frac{e\mu W}{L}[n_0+\frac{1}{\pi }{\left(\frac{\varphi }{hv}\right)}^2],$$(2)where $n_0$ is the residual carrier concentration; $L,W,h,v$, and $\mu$ are the length, width, Bronk constant, velocity, and carrier mobility of the device, respectively. Plugging it into Eq. (1) yields $$S=-\frac{2{\pi }^2{K}_B^{2}T\varphi }{3e[\pi n_0(hv{)}^2+{\varphi }^2]}.$$(3)

Through the above formula, $S$ in Gr can be calculated. When non-centrosymmetric metal is deposited on the surface of Gr, the incident light absorbed by metal is transformed into the surface plasmon resonance, resulting in a significantly enhanced electric field, breaking the original symmetric electric field, and making photogenerated charge carriers gain momentum. In this case, there are generally two ways for carriers to gain momentum. One is caused by the S gradient in Gr, which is mainly based on the ballistic transmission of carriers. The other is guided by the high conductivity of the metal, which is mainly transmitted by carrier diffusion [40]. When the polarized light angle is changed, the photogenerated carrier in the device will obtain different momenta and output different photocurrents through the corresponding vector superposition. The relevant theoretical analysis is shown in Fig. 9. First, consider horizontally polarized light, as shown in Fig. 9(a). The oscillating charge in the horizontal direction of the T-type metal will be excited, while the oscillating charge in the vertical direction will be almost non-oscillating due to the mirror symmetry. Similarly, when the polarization angle is vertical [Fig. 9(b)], the vertical direction of the metal rod generates oscillating charges, while the horizontal direction has almost no charge. Further, when the polarization angle is 45°, it can be mathematically decomposed into polarized light in the direction of 0° and 90°, indicating that it is a vector superposition of the two modes, as shown in Fig. 9(c). Similarly, when the incident polarized light is at any angle, it can be decomposed into two basic modes, and through the corresponding vector superposition, the device can produce different light responses when detecting polarized light at different angles.

Figure 10(a) shows the current-voltage (I-V) curve of the $\mathrm{Gr}/{\mathrm{Al}}_2{\mathrm{O}}_3/\mathrm{GaN}$ heterojunction photodetector under 325 nm ultraviolet light ($0.01-1.5{\mathrm{m}\mathrm{W}\mathrm{cm}}^{-2}$). Obviously, the current of the device at $-2\mathrm{V}$ increases significantly from $1.5\times {10}^{-8}\mathrm{A}$ in the dark to $4.1\times {10}^{-5}\mathrm{A}$ under ultraviolet light ($1.5\mathrm{mW}{\mathrm{cm}}^{-2}$), showing significant photovoltaic behavior.

From the corresponding log I-V curve in Fig. 10(b), it can be observed that with the increase of incident light power density, the photocurrent rises successively but gradually tends to saturation, which is caused by the increase of carrier recombination probability at higher light intensity. Figure 10(c) shows the time-photocurrent response of the device under different light intensities at $-2\mathrm{V}$. It is evident that the device can be reversibly switched between the on and off states stably, with a switching ratio of $2.7\times {10}^3(1.5{\mathrm{mW}\mathrm{cm}}^{-2})$. In addition, there is no obvious current spike in the output photocurrent, indicating that the interface between Gr and GaN is effectively repaired by ${\mathrm{Al}}_2{\mathrm{O}}_3$. Further, according to the optical response characteristics of the device in Fig. 10(d), the corresponding rise time (${\tau }_r$, from 10% to 90% of the peak output photocurrent) and fall time (${\tau }_f$, from 90% to 10% of the peak output photocurrent) are estimated to be 6 ms and 8.7 ms, respectively, at $-2\mathrm{V}$. The related parameters to evaluate the performance of Gr/GaN tunneling heterojunction UV photodetector—responsivity (R), detectivity (${\mathit{D}}^{*}$), linear dynamic range (LDR), external quantum efficiency (EQE), and sensitivity (S)—were calculated and fitted by extracting the data in I-V curve using the following formulas [41–43] [the results are shown in Figs. 10(e)–10(i)]: $$R_{\lambda }=\frac{I_{\mathrm{light}}}{P·S_{\mathrm{eff}}},$$(4)$${\mathit{D}}^{*}=\frac{\sqrt{S_{\mathrm{eff}}}{R}_{\lambda }}{\sqrt{2e{I}_{\mathrm{dark}}}},$$(5)$$\mathrm{LDR}=20\log \left(\frac{I_{\mathrm{light}}}{I_{\mathrm{dark}}}\right),$$(6)$$\mathrm{EQE}=\frac{R_{\lambda }hc}{e\lambda }\times 100\%,$$(7)$$S=\frac{I_{\mathrm{light}}-I_{\mathrm{dark}}}{I_{\mathrm{dark}}}\times 100\%,$$(8)where $I_{\mathrm{light}}$, $I_{\mathrm{dark}}$, $P$, $S_{\mathrm{eff}}$, $e$, and $h$ are, respectively, the output photocurrent of the device, dark current, incident optical power density, effective optical absorption area of the device, electron charge, and Bronk constant. It can be found from the image that with the increase of incident light power density, the $\mathit{R},{\mathit{D}}^{*}$, and EQE of the device decrease, while LDR and S are on the contrary, which is consistent with previous reports [44–49]. At a low incident light power density ($0.01\mathrm{mW}/{\mathrm{cm}}^2$), the device exhibits extremely high R of 97.7 A/W, ${\mathit{D}}^{*}$ of $1.3\times {10}^{14}$ Jones, and EQE of $3.7\times {10}^4$, and the highest LDR and S of the device are 69.3 dB and $2.9\times {10}^5$, respectively, indicating that the device has excellent weak light detection capability, which is very important for specific applications.

The cutting area size with experimental scissors is $0.5\mathrm{cm}\times 0.5\mathrm{cm}$ (with PMMA on the surface) Cu substrate with monolayer graphene; then it was placed in a ferric chloride (${\mathrm{FeCl}}_3,\ge 99.9\%$, Shanghai Aladdin Bio-Chem Technology Co.) solution with a concentration of 1 mol/L for 45 min to remove the Cu substrate at the bottom. The monolayer graphene with the Cu substrate removed was then transferred to a petri dish with deionized water for cleaning to remove the residual ${\mathrm{FeCl}}_3$ solution on the surface. The whole process was repeated 3 times, each time cleaning for 10 min, to achieve the preparation of single layer graphene.

Sapphire of 2-inch size was used as epitaxy substrate, and the heteroepitaxy plane GaN was grown by a two-step method through metal organic chemical vapor deposition (MOCVD). The thickness was 4–5 μm, and the doping concentration was $3.5\times {10}^{17}$ cm⁻³. The detailed preparation process is referred to in our previous work.

First, 1 nm ${\mathrm{Al}}_2{\mathrm{O}}_3$ was deposited on GaN substrate using ALD technology to repair the interface damage. Then, the gallium nitride substrate deposited with the insulating layer was cleaned and used as the transfer substrate for extracting the cleaned monolayer graphene film. It was dried for 1 h and then transferred to the oven at 90°C for 3 h. The thermal evaporation technique was used to deposit different shapes and different numbers of non-centrosymmetric metal structures with a thickness of 100 nm, and the device was immersed in 40°C acetone to remove PMMA. Finally, the metal In was deposited on the GaN and Gr sides for photoelectrical measurements.

In this work, we used FDTD method to obtain the electric field distributions of the nanostructures. During the calculations, we set the $x,y$, and $z$ directions to perfectly match the layer conditions. In addition, we meshed the entire simulation area to $5\mathrm{nm}\times 5\mathrm{nm}\times 5\mathrm{nm}$ to obtain accurate calculation results. Importantly, we employed a total scattered field plane wave incident perpendicularly to the surface of the nanostructures as the excitation light source. Finally, we extracted the electric field distributions of the nanostructures using an electric field monitor.

This study introduces a groundbreaking approach to ultraviolet (UV) polarization-sensitive photodetection by integrating non-centrosymmetric metal nanostructures with a graphene (Gr)/GaN heterojunction. Unlike conventional graphene-based detectors limited by material isotropy and weak UV response, this work leverages asymmetric metal patterns (E-/T-type) to artificially induce directional anisotropy. These structures break symmetry, generating localized plasmonic hotspots and anisotropic electric fields that selectively amplify photogenerated carrier momentum under polarized UV light (325 nm). Coupled with Fowler-Nordheim tunneling (FNT) across an ultrathin ${\mathrm{Al}}_2{\mathrm{O}}_3$ interlayer, the device achieves unprecedented performance: a record anisotropy ratio of 115.5 (E-type, $-2\mathrm{V}$ bias) and an ultrahigh responsivity (97.7 A/W). Critically, the design eliminates reliance on intrinsic material anisotropy or complex nanofabrication, using scalable thermal evaporation for metal patterning. By systematically varying metal geometry and quantity, the study demonstrates tunable, globally enhanced polarization sensitivity—a universal strategy applicable to diverse 2D/3D systems. This innovation bridges the gap between high-performance UV polarization detection and practical manufacturability, opening avenues for advanced optoelectronic applications in imaging, sensing, and optical communications.