Experimental comparison between Nb2O5- and TiO2-based photoconductive and photogating GFET UV detector

In the present study, by adding graphene to a photoconductive photodetector with a niobium pentoxide (Nb2O5) absorber layer and exploiting the photogating effect, the responsivity of the photodetector is significantly improved. In this photodetector, the Nb2O5 layer detects the light, and the graphene improves the responsivity based on the photogating effect. The photocurrent and the percentage ratio of the photocurrent to dark current of the Nb2O5 photogating photodetector are compared with those of the corresponding photoconductive photodetector. Also, the Nb2O5 photoconductive and photogating photodetectors are compared with titanium dioxide (TiO2) photoconductive and photogating photodetectors in terms of responsivity at different applied (drain-source) voltages and gate voltages. The results show that the Nb2O5 photodetectors have better figures of merit (FOMs) in comparison with the TiO2 ones.

www.nature.com/scientificreports/ the lightly-doped Si, electrons accumulate in the traps and create a negative gate voltage, causing more holes to be induced and thus increasing a high gain 25 . In 2018, black phosphorus (BP) with a direct bandgap of 0.3 eV was used as a light-absorbing material. At wavelengths of 655 nm, 785 nm, and 980 nm, the responsivities of 55.75 A/W, 1.82 A/W, and 0.66 A/W were obtained, respectively. Excited electrons are trapped in trap levels, and the holes pass through the graphene layer by the internal potential between graphene and BP. The lifetime of carriers increases with traps. Due to the high mobility of graphene, holes can flow in the circuit before recombining with electrons. The introduced structure works in the visible to near-infrared regions based on the photogating effect 22 .
In 2018, Ti 2 O 3 nanoparticles with a band gap of 0.09 eV were used to fabricate a detector in the mid-infrared spectrum. The mechanism is the same as before. This detector had a responsivity of 300 A/W for a wavelength of 10 μm 28 . In 2018, the photogating effect in graphene photodetectors was investigated using SiO 2 /n-doped Si substrate. At wavelengths of 450 nm and 1064 nm, the responsivities were 500 A/W and 4 A/W, respectively. The bending of the band at the Si/SiO 2 interface separates the electron-hole pairs. Under the electric field, the electrons move towards Si while the holes are trapped at the Si/SiO 2 interface; the accumulation of holes at the Si/Si 2 interface acts like a positive gate and increases the Fermi level of graphene. This causes the graphene to become n-type 24 . In 2018, a graphene transistor was fabricated with an indium antimonide (InSb) substrate. A responsivity of 33.8 A/W with a photogating effect at a wavelength of 4.6 μm was achieved 29 . In recent years various photodetectors with TiO 2 and Nb 2 O 5 absorber layers in the UVA region have been presented 30 . In 2011, an Nb 2 O 5 nanobelt was proposed, and at 1 V, a responsivity of 15.2 A/W was obtained 31 . In 2015, an Nb 2 O 5 nanoplate photodetector was fabricated with a responsivity of 24.7 A/W at 1 V 32 . In 2021, a MAPbI 3 nanowire photodetector was fabricated, and a responsivity of 20.56 A/W at 1 V was reported 33 . In 2023, a type-II heterojunction of TiO 2 NTs/Cs 3 Cu 2 I 5 nanoparticles hybrid fiber was presented with a responsivity of 26.9 mA/W at − 1 V 34 . The photogating effect can be investigated in three different structures, namely quantum dot 10 , bulk [23][24][25] , and thin film structures 8 . Quantum dots can be integrated into two-dimensional materials to achieve some advantages. As the first advantage, quantum dots with greater thickness resolve the issue of the low optical absorption of two-dimensional materials. The second advantage is that two-dimensional materials have high carrier mobility, and the third advantage is that some two-dimensional materials do not have a wide absorption spectrum while quantum dots compensate for this limited responsivity. For a two-dimensional material such as graphene, there is no mechanism to produce multiple carriers from one photon. By using quantum dots, a large number of holes can flow in the circuit, and as a result, the gain increases. This is because the lifetime of the trapped electrons is long, and the mobility of carriers in graphene is high. One of the disadvantages of quantum dots is their toxicity. Moreover, the dimensions of quantum dots change the bandwidth of the used materials. In bulk detectors, due to the defects between SiO 2 and lightly-doped Si, electrons accumulate in the traps and create a negative gate voltage, causing the induction of more holes and thus increasing the gain. In other words, the bending of the band at the Si/SiO 2 interface separates the electron-hole pairs. Under the internal field, the electrons move towards the Si substrate while the holes are trapped at the Si/SiO 2 interface, and the accumulation of holes at the Si/SiO 2 interface acts like a positive gate and increases the Fermi level of graphene. As a result, the graphene is converted into n-type graphene. A highly-doped silicon substrate is not used since it has additional carriers with a much shorter lifetime. The application of the bulk structure is limited to high-energy materials and X-rays [23][24][25] .
In the present study, Nb 2 O 5 (3.7 eV) and TiO 2 (3.2 eV) absorber thin films are used based on the photogating mechanism. The use of broadband materials is an advantage for the photodetector because it operates at room temperature. By transferring graphene to the photoconductive detector, the responsivity increases approximately 20 times. The use of an Nb 2 O 5 absorber layer is a new technique. The advantages of this technique are its low cost and simplicity of obtaining Nb 2 O 5 without any special technology only by combining a few solutions. In the present study, a photoconductive detector with an Nb 2 O 5 absorber layer and a graphene photogating photodetector with the same Nb 2 O 5 absorber layer are compared in terms of responsivity and the percentage ratio of photocurrent to dark current. The responsivities of the photogating and photoconductive detectors with the Nb 2 O 5 absorber layer are 12.69 A/W and 0.65 A/W respectively. The percentage ratios of photocurrent to dark current of the photogating and photoconductive detectors with an Nb 2 O 5 absorber layer are respectively 2.84% and 0.16% at a drain-source voltage of 1.5 V and a gate voltage of 1 V. The responsivities of the TiO 2 -based photoconductive and photogating detectors, which are fabricated under the same laboratory conditions, are 0.45 A/W and 8.32 A/W, respectively. The percentage ratios of the photocurrent to dark current of the TiO 2 photoconductive and photogating detectors are 0.16% and 2.84%, respectively. The responsivity of a photogating detector with an Nb 2 O 5 absorber layer is about two times higher than that of a TiO 2 photogating detector.

Fabrication process
The fabrication steps of the photoconductive and photogating photodetectors with an Nb 2 O 5 absorber layer are shown in Fig. 1a-g, respectively. Figure 1a shows a p-type silicon wafer with a thickness of 430 μm along the (100) direction. The silicon surface was cleaned by the RCA method. As shown in Fig. 1b, an oxide layer was formed on silicon with 300 nm thickness using thermal oxidation. The doping amount was 11-13 Ω/cm, and the leakage current density of the oxide sample was 0.205 A/m 2 . As shown in Fig. 1c, 30 nm Nb was deposited on the SiO 2 layer by electron-beam physical vapor deposition. Then, as shown in Fig. 1d, an anodic process was used to form an 81-nm Nb 2 O 5 absorber layer on SiO 2 . For the anodization of Nb, an electrolyte consisting of 1200 ml of ethylene glycol C 2 H 6 O 2 , 800 ml of H 2 O, and 160 g of (NH 4 ) B 5 O 6 was used. The Nb layer was connected to the positive pole and the platinum electrode, which was inside the anodization solution, was connected to the negative pole, and the anodization process was carried out. As shown in Fig. 1e,f, the interdigit electrodes were patterned on the structure using the lithography process. For this purpose, a glass mask was used. The width of each metal electrode was 10 μm, and the width of the transistor (w) was 5000 μm. The distance between two www.nature.com/scientificreports/ metal electrodes (length) was 12.5 μm, and the number of the distances (n) was 249. Accordingly, as shown in Eq. (1), the total width of the transistor is 1245 mm.
where W Total is the total width of the transistor, n is the number of distances, and w is the width of the transistor.
To fabricate the photogating photodetector, all the steps are the same as performed for the photoconductive photodetector; only at the end, as shown in Fig. 1g, graphene is transferred to the final photoconductive structure. In this structure, graphene, which was the product of the Graphene Company and grown using the chemical vapor deposition (CVD) method, was used 35,36 . Figure 2 shows the energy bands for graphene and Nb 2 O 5 heterostructures. By placing the graphene monolayer on the Nb 2 O 5 layer, due to the difference in the Fermi level of the two materials, a built-in potential barrier is created between the two materials. Then, light radiation in the range of the band gap of the Nb 2 O 5 layer leads to electron-hole pair production inside the Nb 2 O 5 layer. Due to the potential barrier between the graphene and Nb 2 O 5 layer, the electrons move towards the graphene layer, but the holes are trapped inside the Nb 2 O 5 layer. The trapping of the holes changes the Fermi energy level of graphene and the resistance of the graphene channel, causing a large photocurrent [22][23][24][25][26] .
Graphene was transferred through a wet transfer method in seven stages, including etching the copper layer, substrate preparation, and graphene transferring. The process begins by eliminating the unwanted graphene that is placed under the copper layer during the CVD process. This process was done in a 20% nitric acid solution in which the substrate was kept for about five minutes. Then, the copper was etched using 0.2 M iron (III) chloride for almost two hours. After the metal residues were removed by RCA, the substrate was ready for graphene transferring, but before that, the substrate must be prepared. Next, the substrate was immersed in piranha solution at 70 • C for about 15 min. Then, graphene was transferred to the prepared substrate. The following two stages were performed to increase the adhesion at the graphene/substrate interface and remove poly(methyl methacrylate) (PMMA). The substrate was heated moderately using a heater to a temperature of 80 • C for about five minutes before being subjected to a higher temperature of 130 • C for approximately 20 min. Next, the substrate was allowed to cool down for a few minutes at room temperature. The process finished by removing PMMA via an N-Methyl-2-pyrrolidone (NMP) solution at a temperature of 70 • C for about 15 min. Figure 3 shows the scanning electron microscopy (SEM, Tescan VEGA3) images of the graphene transferred onto the SiO 2 layer and interdigit electrodes to the Nb 2 O 5 layer. Figure 3a shows the grains of the monolayer graphene that was annealed for one hour at 550 ℃ and a vacuum of 4.4 × 10 -6 Torr to remove the PMMA residues. The average graphene sheet resistance was obtained by the van der Pauw method. An HP4450 semiconductor measurement device was used at a voltage of 1 V and a current of 1 μ A . The sheet resistance of the graphene layer was around 1600 ±10% ohms/square. An example of the interdigit electrode design is shown in Fig. 3b. The contact resistance of graphene and nickel electrodes were measured to be about 640 Ω μm ± 15% using the transmission line method (TLM). Fortunately, the contact resistance was good enough such that it did not affect the process of fabricating the photodetector although a lower contact resistance would be better.

Equations
In the photogating effect, if one species of the generated carriers is trapped, it can generate an additional electric field, in the form of the gate voltage, to modulate the channel conductance. Photodetectors with small dimensions show high responsivity. In the photovoltaic effect, the produced electron-hole pairs are recombined in picoseconds. However, the carrier lifetime in the photogating effect is longer than that in the photovoltaic effect, i.e., 1 s compared to around 20 ms 27 . The gain can be obtained using Eq. (2). As can be concluded from this equation, responsivity and photocurrent increase with the increase in the trapping time and the decrease in the  www.nature.com/scientificreports/ time interval during which carriers pass through the graphene channel (transit time). As the graphene channel becomes shorter, the photocurrent and responsivity increase, while the effective area of the photodetector decreases 38 .
The time during which free charge carriers travel from drain to source is denoted by "τ tr " and the time interval during which these free charge carriers are trapped near nanoparticles is denoted by "τ life ".
If the lifetime of an extra electron is longer than the transit time (τ life > τ tr ), the extra electron reaches the anode, and another electron immediately enters the photoconductor to maintain the charge neutrality and drifts to the anode terminal. This process is repeated until the extra electron recombines with a hole. This process takes τ life on average, and the gain is greater than unity. However, if τ life < τ tr , the extra electron recombines with a hole before the transit is completed. To achieve a gain of greater than unity without multiple electron-hole pair production, a higher power by an external circuit is needed. Equation (3) calculates the photocurrent of the photodetector 39 .
where C ox is the dielectric capacitor per unit area, and W and L are respectively the width and length of the graphene channel.
From Eq. (4), the net photocurrent is independent of the thickness of the SiO 2 film, while it depends on the carrier transit time and the amount of photoinduced electric charge as follows: where ∆Q is the amount of photoinduced electric charge, and τ tr is the transit time of the carrier in the graphene channel.
As shown in Eq. (5), the amount of current generated in this type of detector depends on α and d as given below 40 .
where α is the absorption coefficient, and d is the absorber layer thickness.
To increase the current, materials with a high absorption coefficient are used because according to Eq. (6), with increasing d, C ox decreases. Moreover, according to Eqs. (7) and (8), the decrease in C ox reduces g m , and as a result, photocurrent decreases. Therefore, the absorber layer thickness, d, should have an optimal limit, which is equal to 81 nm (Nb 2 O 5 ) in the present study. The oxide-to-metal conversion factor of Nb is 2.7, meaning that for 1 nm of Nb, 2.7 nm Nb 2 O 5 is formed.
where ε is the dielectric constant, and ε 0 is the vacuum permittivity.
where g m is the transconductance given as follows: where V GS is the control voltage.

Results and discussion
Some of the figures of merit (FOMs) of the photodetectors are evaluated optically. The optical characteristics, except for some special cases, were obtained using a helium-cadmium laser at a power of 1 μW, a wavelength of 325 nm, and a chopper frequency of 3 kHz. As shown in Fig. 4, for optical characterization, the laser light hits a mirror, and the reflected light reaches a chopper and then the photodetector. The detector output is connected to the input of a lock-in amplifier.
Using this optical setup has two advantages. The first advantage is that the ambient noise is eliminated, and the second advantage is that even the lowest amount of photocurrent can be observed with this setup. The photoconductive and photogating photodetectors with Nb 2 O 5 and TiO 2 absorber layers were optically examined.
As shown in Table 1, the Nb layer was deposited on the silicon substrate by an electron beam device, and the Nb 2 O 5 layer was obtained by anodizing with an oxide-to-metal conversion factor of 2.7. The Ti layer was deposited on the silicon substrate by sputtering, and TiO 2 was obtained by the oxidation method with an oxideto-metal conversion factor of 1.7. Figure 5a shows the profile of the Nb 2 O 5 photoconductive detector, and Fig. 5b shows the profile of the Nb 2 O 5 photogating detector. As shown in Fig. 5, the drain-source voltage (V DS ) is applied to both ends of the graphene layer, and the control voltage (V GS ) is applied to the silicon substrate through ohmic contacts.
(3) www.nature.com/scientificreports/ Figure 6a shows the photocurrent in μA versus drain-source voltage for the photoconductive and photogating photodetectors with the Nb 2 O 5 absorber layer. The photocurrent of the photoconductive photodetector increases from 0.533 to 0.64 μA. By adding graphene, the photocurrent increases from 7.87 and 12.69 μA in the drainsource voltage range of 0.1-1.5 V. As the value of V DS increases, due to the reduction in the potential barrier, electrons more easily travel from Nb 2 O 5 to graphene, and as a result, the increase in I ph is greater. In other words, according to Eq. (6), with the increase in V DS , g m and as a result, I ph increases. Figure 6b shows the percentage ratio of photocurrent to dark current for the photoconductive and photogating photodetectors with the Nb 2 O 5 absorber layer. With increasing drain-source voltage, the ratio of photocurrent to dark current decreases, and the lowest value is observed at the drain-source voltage with the highest responsivity, i.e., at 1.5 V. Not only responsivity changes slightly with the increase in voltage, but also a further increase in V DS allows much current to pass through the photodetector, and the device heats up. This has an adverse effect on the performance of the device. As shown in Fig. 6a, the value of I ph has a direct relation with V DS , while I dark depends not only on V DS but also on other factors, such as carrier lifetime. Therefore, the trend of the I ph /I dark curve results from the differences between the slopes of I ph and I dark . In the photoconductive and photogating detectors, the percentage ratios of photocurrent to dark current are 0.16% and 2.84%, respectively, at a wavelength of 325 nm, a V DS of 1.5 V, and a power of 1 μW.   www.nature.com/scientificreports/ A simple criterion is defined to compare the dark current of photoconductive and photogating detectors at different drain-source voltages. In Fig. 7, the ratios of the dark current of the detectors are compared. This figure shows the ratio of the dark current of the photoconductive photodetector with Nb 2 O 5 to that with TiO 2 and the ratio of dark current of the photogating photodetector with Nb 2 O 5 to that with TiO 2 , versus drain-source voltage. In general, the ratio of the dark current with the Nb 2 O 5 absorber layer to the dark current with the TiO 2 absorber layer is higher in the photogating detector than in the photoconductive detector. The dark currents of the photogating and photoconductive detectors with the Nb 2 O 5 layer are better than those of the photogating and photoconductive detectors with the TiO 2 absorber layer. Figure 8 shows the responsivities of four photoconductive and photogating photodetectors with Nb 2 O 5 and TiO 2 absorber layers in the V DS range of 0.1-1.5 V. The photogating photodetector with the Nb 2 O 5 absorber layer has the highest responsivity at all V DS voltages. Moreover, the responsivities of the photoconductive detectors do not change significantly with increasing V DS unlike the two photogating photodetectors. However, due to the photogating effect of the two detectors (i.e., detectors with Nb 2 O 5 /Gr and TiO 2 /Gr layers), the carrier trapping time increases, and thus, the responsivity and photocurrent increase. The responsivity is also dependent on the initial energy level of the graphene layer. At a drain-source voltage of 1.5 V, the ratio of responsivity of  www.nature.com/scientificreports/ the Nb 2 O 5 /Gr photodetector to that of the TiO 2 /Gr photodetector is about 2, and the ratio of the responsivity of the Nb 2 O 5 photodetector to that of the TiO 2 photodetector is about 1.5. This indicates the better performance of the Nb 2 O 5 absorber layer compared to the TiO 2 absorber layer while both of these layers were made under the same laboratory conditions. As shown in Table 2, the responsivities of the photogating detectors are better than those of the photoconductive detectors. The responsivities of both types of detectors with the Nb 2 O 5 absorber layer are higher than those with the TiO 2 absorber layer. The value of I Nb 2 O 5 /Gr dark /I TiO 2 /Gr dark of the photogating detectors is larger than the value of I Nb 2 O 5 dark /I TiO 2 dark of the photoconductive detector. Figure 9a shows the responsivity in terms of gate voltage. As shown in this figure, the responsivities of the photoconductive detectors do not change significantly with the increase in V GS . However, in the photogating detectors, the carrier trapping time increases due to the photogating effect, so responsivity and photocurrent increase. The responsivity is also dependent on the initial energy level of the graphene layer. Therefore, it is proven that the increase in the responsivity of the photodetector with increasing V GS is because of transferring the graphene layer to the TiO 2 absorber layer.
As shown in Fig. 9a, with increasing gate voltage, the responsivity of the photogating detectors increases. The Dirac points of the Nb 2 O 5 and TiO 2 photogating detectors are located at 20 V and 15 V, respectively. At the Dirac points, not only the photogating detectors have the highest responsivity, but also the curves have the highest slope around the Dirac points. As shown in the curves of the photogating detectors, further increase in the gate voltage has a downward trend, and responsivity decreases. The reason for this is shown in Fig. 9b. In this figure, the blue curve shows the photocurrent in terms of the gate voltage in the graphene material. The minimum of the blue curve is the Dirac point. The Dirac voltage corresponds to when the numbers of electrons and holes are the same. As it is clear from the blue curve, at the gate voltages that are far away from the Dirac point, the photocurrent is independent of the gate voltage, so increasing or decreasing the gate voltage does not affect the value of photocurrent. As shown, the steepest slope of the curve is somewhere near the Dirac point. From Eq. (8) and the green curve, this slope is at gm = ∂I d /∂V g , and the maximum of the responsivity curve, which is shown in red, is located at the points with the highest slope.  www.nature.com/scientificreports/ As shown in Fig. 10, the bandwidth of the photoconductive and photogating photodetectors with an Nb 2 O 5 absorber layer was calculated as an FOM. With the increase in frequency, the frequency response criterion does not decrease substantially. These results indicate that both photodetectors can be used in applications, such as UV imaging and many modern systems, with a frequency requirement of up to 5 kHz. Table 3 compares the photoconductive/photogating detectors with the Nb 2 O 5 /TiO 2 absorber layers (presented in this study) and previous detectors presented in other articles with the same absorber layer materials.

Conclusion
An Nb 2 O 5 absorber layer was used for fabricating a graphene detector based on the photogating effect. Moreover, an anodizing process was employed to produce this absorber layer from a thin layer of Nb. These processes have not been conducted in the literature and were presented in this study for the first time. It is worth noting that compared to photodetectors with a TiO 2 absorber layer, the photodetectors with an Nb 2 O 5 absorber layer performed better in terms of the cost, simplicity of the fabrication process, and high value of responsivity. With 325-nm laser radiation, 1 μW power, at a drain-source voltage of 1.5 V, and a gate voltage of 1 V, the responsivities of the photoconductive and photogating detectors with the TiO 2 absorber layer were 0.45 A/W and 8.32 A/W, respectively, while the responsivities of the photoconductive and photogating detectors with the Nb 2 O 5 absorber layer were 0.65 A/W and 12.69 A/W, respectively. Moreover, the percentage ratios of the photocurrent to dark current of the photoconductive and photogating detectors with the TiO 2 absorber layer were 0.003% and 0.111%, respectively. However, the percentage ratios of the photocurrent to dark current of the photoconductive and photogating detectors with the TiO 2 absorber layer were 0.16% and 2.84%, respectively. Therefore, the responsivities and the percentage ratios of the photocurrent to dark current in detectors with the Nb 2 O 5 absorber layer were better than the detectors with the TiO 2 absorber layer.

Data availability
The datasets used and analyzed during the current study are available from the corresponding author upon reasonable request.