Introduction

Photoelectric conversion, with its broad range of applications, is at the center of numerous technological and scientific fields, including imaging, optical communication, solar energy, energy harvesting1, remote sensing, photoelectric memories, environmental monitoring, military reconnaissance, and astronomical studies2,3.

In recent years, one of the most promising materials is represented by graphene (Gr) which has attracted enormous attention owing to its 2D characteristic properties that impact on electrical, mechanical, and thermal properties, making it useful for variety of applications such as solar-cells, field effect transistors, sensors, and photodetectors3,4,5. More specifically, heterostructures involving 2D materials when joint a 3D semiconductor, i.e., silicon, allow achieving heterointerfaces having the optical properties of dissimilar materials6,7,8,9, facilitating the band alignment and optical transactions between them, improving the effective spatial separation of electron–hole pairs and tunneling efficiency10,11.

Moreover, the compatibility of graphene as a material with well-established silicon (Si) fabrication processes makes it a promising candidate for large-scale and low-cost technological applications12,13,14. Indeed, there is a growing effort in developing high-performance graphene/Si photodetectors15,16,17, in which the photoexcitation resides in Si, while the graphene plays the role of a highly efficient carrier collector due to its low light absorbance (2.3%)18. To avoid the pinning of the Graphene Fermi level by the silicon, which induces a current leakage that limits the photoresponse, an insulating layer is introduced between the graphene and silicon. In this way, a graphene/insulator/Si heterojunction is realized. In this structure, it is possible to reduce the pinning of the Graphene Fermi energy level of graphene by applying a reverse bias19,20. This feature addresses a key issue in enhancing the capture of photoexcited carriers, resulting in high photocurrent responsivity.

In this paper, the heterojunction made of a graphene monolayer on Si3N4/n-silicon substrate, in which graphene acts as a metal-like material, connected to a Schottky junction (n-Si/metal) on the backside, is presented. This device structure enables the reduction of dark leakage current and an improvement in photoresponse performance in the UV-NIR range compared to the same graphene/Si3N4/n-silicon heterojunction connected to an Ohmic contact on the backside. Furthermore, a detailed analysis of the mechanism involved in the carriers’ transport process is proposed. The findings suggest that the realized device offers competitive performance, thanks to the heterojunction structure combined with the tunability of the graphene Fermi level. This work provides an understanding of the features useful to improve the photoresponse in graphene/silicon-based devices, enabling high-performance, low-power, and low-cost photodetectors for future optoelectronics.

Results

A single layer of graphene (SLGr) has been transferred onto the surface of n-silicon-based substrate to realize graphene-based heterojunction photodetectors. In Fig. 1, the cross-section structure of the two fabricated devices is shown: the surface of the device is made by a graphene monolayer, which covers the Si3N4/n-silicon substrate. Considering graphene as a metal-like material, the junction SLGr/Si3N4/n-S from now on will be identified as an MIS (metal-insulator-semiconductor) junction. On the rear side, one device has an Ohmic contact (Fig. 1 (a)), and the other has a Schottky contact (Fig. 1 (b)). Further information on the device structure is given in the Method section. To simplify the notation, the device consisting of the MIS junction coupled to the Ohmic contact and the Schottky contact on the rear side will be referred to as the Ohmic device and the Schottky device, respectively.

Fig. 1
figure 1

Schematic representation of heterojunction-based devices: SLGr/Si3Ni4/n-Si (MIS) junction with (a) Ohmic contact and (b) Schottky contact on the backside. The dashed lines near the MIS and Schottky interfaces represent the depletion regions. The electrical connection in the figure shows the forward bias configuration; (c) optical image of the SLGr/Si3Ni4/n-Si device.

Full size image

A preliminary analysis of the current-voltage (I-V) response in the dark and under illumination (at a wavelength of 600 nm) of the graphene single layer transferred to the top of the substrate has been carried out (see Fig. 2(a)). The I-V characteristics of SLGr were measured using only the two Pt/Ti pads on the top surface. The I-V curves, from − 10 V to 10 V reported in Fig. 2 (a), reveal that the current under illumination is slightly higher compared to the dark current. This result confirms that the photocarrier generation is unlikely to occur in the graphene layer due to the ultrafast recombination in graphene and its low absorbance (about 2.3%) under illumination21,22. Otherwise, the linear behavior and the resistance of 15 kΩ suggest the formation of a very low contact resistance between the Pt/Ti pads and the graphene monolayer.

Fig. 2
figure 2

(a) I-V characteristics in the dark and under illumination condition (at 600 nm with a power of 0.5 mW) of (a) the graphene monolayer and (b) graphene/n-Si photodetector with Ohmic contact on the backside. The positive (negative) voltage corresponds to the forward (reverse) bias.; (c) photocurrent at wavelengths ranging from 400 nm to 800 nm for the Ohmic device as function of reverse bias voltage (applying negative voltage on the graphene pads and positive voltage on the back side of the device).

Full size image

After this preliminary analysis, the I-V curves of the Ohmic device were measured in the dark and under illumination (at a wavelength of 600 nm), applying the bias using the connection shown in Fig. 1.

From Fig. 2 (b), the device exhibits clear rectifying behavior, both in the dark and under light, with an apparent photodetection effect. In particular, in the reverse bias (negative voltage), the photocurrent is much higher than the dark current. In contrast, in the forward bias, there is no difference between the dark current and photocurrent. This result demonstrates that the device based on a graphene/n-type silicon substrate exhibits photogenerated charges, mainly when the device is in reverse bias.

To study how the reverse bias current depends on wavelength of the light in the range 400–800 nm, the I-V of the Ohmic device was recorded and shown in Fig. 2 (c). The curves have the same upward trend that does not depend on the light wavelength, while the current intensity changes with the incident light, and it reaches a peak corresponding to the wavelength of 680 nm. In all the curves, the current increases sharply, starting from a common voltage threshold (Vth) of approximately 18 V (see the arrow in Fig. 2(c)). It then does not saturate but instead increases gradually with the applied voltage. With the aim to overcome this feature, i.e. to stabilize the current and to minimize the leakage to obtain a current saturation regime typical of a Schottky diode, the substrate structure has been modified: the Ohmic contact on the rear side has been replaced by a Schottky contact, as shown in Fig. 1 (b).

Fig. 3
figure 3

(a) the I-V characteristics of the SLGr/n-Si photodetector with Schottky contact on the rear side in the dark and under illumination (wavelength of 600 nm); (b) Photocurrent versus voltage bias of the Schottky device at different light wavelengths ranging from 400 nm to 800 nm. The arrow VTh indicates the threshold voltage.

Full size image

The current-voltage characteristic of the Schottky device (Fig. 3 (a)) reveals a significant rectification behavior in the reverse bias condition, where a substantial photoresponse appears. Indeed, the graph shows that a lower dark current (a few nanoamperes) is obtained in the Schottky device compared to the Ohmic device (see Fig. 2 (b) negative bias voltage).

A closer look at the current of the Schottky device under reverse bias, in both dark and light conditions (Fig. 3(b)), reveals some differences compared to the Ohmic device in terms of trends and threshold voltage. As reported in Fig. 3 (b), the current of the Schottky device experiences a threshold Vth around 11 V after which it increases sharply before leveling off in a saturation regime. Otherwise, by introducing the Schottky contact, two features characterize the I-V curves of the device: the threshold (VTh) at 11 V (indicated by the arrow in Fig. 3(b)) and the presence of a saturation current.

It was observed that under illumination, the photocurrent begins to saturate at different bias voltages with a wavelength dependence. It is reasonable to assume that the bias voltage leading to the saturation current depends on the photogenerated charges within the silicon depletion layer, which increase with increasing wavelength up to 680 nm light source (the penetration depth of the radiation in the silicon increases from UV to Vis range)11. In other words, the bias voltage that produces the saturation current is the voltage required to drain all the photogenerated charges to the electrodes. Therefore, in the NIR region (> 700 nm), the photogenerated charges decrease (high transmittance of the silicon), and consequently, the saturation current occurs at a lower bias voltage. Furthermore, the estimated current value in the saturation regime shows a linear dependence on the light powers down to a few µW (see Supplementary Information S1).

Fig. 4
figure 4

Figures of merit (a) Responsivity, (b) Detectivity, and NEP versus wavelength of the Gr/n-Si device with Schottky contact.

Full size image

In order to consider the Schottky device as a promising photodetector, an estimation of the responsivity, which is one of the most important figures of merit for a photodetector, has been calculated. Figure 4 (a) shows the responsivity as a function of the illumination wavelengths for the Schottky device in the UV-Vis-NIR region, evaluated using the relation \(\:R=\frac{{I}_{ph}}{P}\) where Iph is the photocurrent extracted at 28 V (saturation regime), and P is the light power. The graph shows a maximum responsivity of 1.1 A/W at 680 nm, while at 400 nm and 800 nm, values of about 0.3 A/W and 0.8 A/W were measured, respectively. Furthermore, since the photocurrent does not change when moving the light spot on the graphene surface, the device has the same responsivity over a large area (see Supplementary Information S2). These findings could be attributed to the high-quality structure with a large back electrode, as well as the presence of graphene, which efficiently collects the photogenerated carriers.

In addition to the responsivity, other key performance parameters have been evaluated: the specific detectivity \(\:{D}^{*}\) (\(\:{D}^{*}=\frac{R}{\sqrt{2e{J}_{d}}}\) where R, e, and Jd are the responsivity, elementary charge, and dark current, respectively) and the noise equivalent power (NEP), which is related to the D* by \(\:NEP=\frac{\sqrt{A}}{{D}^{*}}\) (A is the active area of the device). Figure 4 (b) reports the D* and the NEP of the Schottky device as a function of the wavelengths. It could be noted that at 685 nm there is a maximum of specific detectivity (~ 3 × 1012 Jones) and a minimum of NEP (20 × 10−14 W/Hz1/2). Thanks to the achieved performance parameters, the realized Schottky device outperforms previous graphene-based heterojunction photodetectors and yields results comparable to or better than those of silicon-based commercial photodetectors typically used in the conventional triode configuration11,23,24,25,26,27.

Discussion

The results reported in the previous section demonstrate that the fabricated photodetectors exhibit better performance when the graphene single-layer-based heterojunction is connected to the Schottky contact on the backside (named Schottky device).

As mentioned before, the Schottky device has a vertical structure consisting of two junctions: a MIS junction, made of the SLGr/Si3N4/n-Si, followed by a Schottky contact formed by the semiconductor/metal (n-Si/(Pt/Ti)) junction, as shown in Fig. 1 (b). The coexistence of these two junctions induces the formation of the depletion regions (dashed lines in Fig. 1 (b)), creating across the heterojunction the built-in potential that, together with the reverse bias voltage, rapidly separates the photogenerated electron-hole (e-h) pairs that can flow through the external circuit.

In the Schottky device, the graphene, directly exposed to illumination, acts mainly as an optically transparent and charge collector20 since it has a weak absorption (2.3%) and its photogenerated carriers have a short lifetime of the order of a few picoseconds. The MIS heterojunction is characterized also by the presence of a thick insulator (60 nm) made of Si3N4 layer to increase the Schottky barrier between the direct contact of graphene and silicon. As the Si3N4 is chemically stable and easy to grow on Si, it is well accepted in Si device technology, and it represents a strong candidate material for interfacial insulators28. Moreover, the presence of nitrate reduces the Fermi level pinning29,30 between the graphene and the silicon, reducing the leakage current and allowing the device to break down at high reverse voltages31.

Fig. 5
figure 5

Energy band diagrams of a MIS junction (SLGR/Si3N4/n-Si) at zero bias, V = 0 (a); and applying a negative bias to the graphene monolayer: (b) V < 0 depletion conditions and (c) V < < 0 in inversion conditions. The graphene single layer is p-doped so that the Fermi energy level is located below the Dirac point. The striped zone represents hole states.

Full size image

The photodetection performance of the realized device depends on the transport mechanism that occurs when the MIS heterojunction is reverse-biased. As reported in Fig. 5, the MIS energy band diagram has two possible configurations as a function of the reverse applied voltage: depletion and inversion state, V < 0 (Fig. 5 (b)) and V < < 0 (Fig. 5(c)), respectively32.

It could be noted that at zero bias (Fig. 5(a)) the bands are flat and no current flows in the device, while applying a reverse bias (the pad in contact with the graphene is negatively biased) the bands are bent upward, the semiconductor surface is depleted, so that a depletion region appears, as shown in Fig. 5 (c).

Due to the band bending in the inversion state, the recombination rate is strongly reduced due to the accumulation of holes and reduction of the electron concentration at the interface Si3N4/n-Si. At the same time, the Fermi level of the graphene is shifted with respect to the bulk silicon band by the reverse bias, opening a larger number of accessible states for the hole injection20,33. As reported in the band diagram, since the graphene is p-doped, due to the ambient exposure as demonstrated in34, the Fermi energy level of the graphene is shifted below the Dirac point.

Then, in the inversion regime, the accumulation of holes at the interface favors hole tunneling through the nitride into the graphene; tunneling dominates recombination, and the current suddenly increases above a specific threshold bias voltage (VTh), which can be considered the device switching bias.

As reported in Fig. 2(c) and Fig. 3 (b), the I-V curves show some differences between the two devices in terms of the threshold bias and the trend of the saturation current. These features are necessarily dependent on the inversion state of the MIS junction and its coupling to the junction on the backside of each device. Therefore, a detailed understanding of the current transport mechanisms is required, so that the graph of the dark current versus voltage on a double logarithmic scale is reported in Fig. 6. This graph allows the identification of the current power dependence on the applied voltage, which characterizes the different transport mechanisms.

For the bias voltage lower than VTh, in the Ohmic device the thermionic (TE) mechanism dominates (the slope is equal to 0.5). In contrast, in the Schottky device, at very low voltage and up to 4 V, the Ohmic mechanism appears (slope is about 1) followed by the TE mechanism (slope is equal to 0.5) that persists just before the threshold voltage VTh33,35. This behavior could be ascribed to the contact resistance due to the presence of the Schottky contact on the back side, which at very low voltage is present32, while the TE process appears at voltages higher than 4 V and persists up to the VTh, as well as for the Ohmic device.

As the bias voltage is increased for the V> VTh, the I-V curves are characterized by a knee, after which a sudden linear increase in current appears, which results in good agreement with the fitting curve of the Fowler-Nordheim (FN) models (blue lines in Fig. 6(a)). See Supplementary Information S3 for more details on the analysis and plots of the FN fits.

Fig. 6
figure 6

(a) Current-Voltage characteristics of devices with Schottky junction (empty circles) and with Ohmic contact (red circles) on the backside. For each curve, the transport mechanism involved is indicated as a function of the applied voltage. (b) and (c) Schematic band diagram of the two devices: with Schottky and with Ohmic contact on the backside, respectively. TE and FN represent the thermionic and Fowler-Nordheim mechanism; EF is the Fermi level (the superscript GR and Si are referred to the graphene monolayer and Si, respectively); \(\:{\varphi\:}_{B}^{MIS}\) and \(\:{\varphi\:}_{B}^{SC}\) are the Schottky barrier of the MIS and Schottky junction; The band diagrams (b and c) are referred to the configuration in which the pad above the graphene monolayer is connected to ground, while the Pt/Ti pad on the backside is positively biased.

Full size image

In general, the FN tunneling is characterized by the barrier having a triangular shape and tunnelling through only part of the insulator. In the case of a thicker insulator layer, the FN current depends mainly on the applied electric field33. Then, the FN mechanism occurs when the applied electric field is large enough to produce the narrowing of the barrier, allowing the electrons to cross the potential barrier into the conduction band of the dielectric. Since the FN models fit the experimental data, it is reasonable to consider that this mechanism is related to the tunneling of electrons through the nitride layer toward the silicon. From the slope of the FN linear fit, the tunnel barrier across the MIS junction was extracted to be 0.87 eV and 0.6 eV for the Schottky and Ohmic devices, respectively36.

Here, the different values of the barrier suggest that in the two devices, the graphene Fermi level is located at other positions with respect to the silicon bands (Fig. 6(b) and (c)). Since the graphene Fermi level is bias dependent37,38, applying a negative voltage to the graphene, the upward shift of its Fermi level opens hole states that could be occupied by the charges incoming from the silicon. Furthermore, when the ohmic contact is present on the back side of the device, a lot of electrons flow across the back pad, enabling a considerable number of holes across the nitride/silicon interface. In this way, in the graphene, a lot of hole states are occupied by the charges that come from silicon, producing a greater upward shift of the graphene Fermi level, reducing the barrier height. On the other hand, the presence of the Schottky contact on the back side reduces the electron flow and consequently the number of holes across the nitride/silicon interface. Then, in the graphene, only restricted hole states could be occupied, reducing the upward shift of the Fermi level and consequently the current flow. Therefore, it is reasonable to assert that the graphene Fermi level is tuned by both the negative bias and the junction on the back side of the device. The different barrier height extracted by the FN model implies that the presence of the Schottky junction pinned the graphene Fermi level much more than the ohmic contact.

In addition, a closer look at the plot in Fig. 6 (a) reveals another difference between the Schottky and Ohmic devices in the current-voltage plot. In the Schottky device, the current rises sharply for V > VTh and then levels off even as the voltage increases (black dashed line in Fig. 6 (a)). Since the current does not depend on the applied electric field in this range, it is reasonable to assume that the electrodes collect all the charges involved in the conduction process, and a steady flow of charges passes inside the external circuit. Furthermore, this feature can also be attributed to the presence of series resistance introduced by the Schottky junction on the back side, which limits the current, as typically occurs in conventional rectifying photodetectors39.

The analysis of the transport mechanisms suggests that in the devices the TE mechanism affects significantly the conduction. The implication of this mechanism is supported by the comparison of the thermal energy with the parameter defined as \(\:{E}_{00}=\frac{qh}{2}\sqrt{\frac{N}{{m}^{*}{\epsilon\:}_{s}}}\), where q, h, N, m*and εs is the elementary charge, the Planck constant, the doping concentration of the substrate, the effective mass of the electrons and the silicon dielectric constant, respectively33,41. In our case, the n-silicon substrate has N≈1012 cm−3, which implies the \(\:{E}_{00}=\:\text{1,7}\times{10}^{-6}\:eV,\) consequently \(\:{E}_{00}/KT<0.2\) (K is the Boltzmann constant and T is the absolute temperature), which means that the thermionic mechanism plays an important role in the conduction process40,41.

Therefore, considering the TE mechanism, the Schottky device could be described as two Schottky barriers (MIS and Schottky junctions) connected back-to-back (Fig. 7(a)). As reported in literature42, the current due to two Schottky barriers could be described as:

$$\:{I}_{T}\propto\:\frac{2{I}_{SC}\cdot{I}_{MIS\:}sinh\left[exp\:\left(-\frac{q{V}_{SC}}{2kT}\right)\:\right]}{{I}_{SC\:}exp\:\left(-\frac{q{V}_{SC}}{2kT}\right)\:+{I}_{MIS\:}exp\:\left(-\frac{q{V}_{SC}}{2kT}\right)\:}$$
(1)

with \(\:{I}_{SC,MIS}={\pm\:I}_{0}\left[exp\:\left(\pm\frac{\:q{V}_{SC,MIS}}{\eta\:kT}\right)\:-1\right]\), where the saturation current is \(\:{I}_{0}={A}^{*}{T}^{2}exp\:(-\sqrt{\chi\:}\:\delta\:)\:exp\:\left(-\frac{q{\varphi\:}_{B}^{SC,MIS}}{kT}\right)\), in which A* is the effective Richardson constant, T is the absolute temperature, \(\:\chi\:\) is the mean barrier of the nitride, \(\:\delta\:\:\)is the thickness of the nitride, \(\:{\varphi\:}_{B}\) is the Schottky barrier with the notation SC and MIS referred to the Schottky and MIS junction, respectively.

Fig. 7
figure 7

(a) The equivalent circuit of the photodetector device structure made by two back-to-back junctions at reverse bias configuration: comparison between the experimental data and the fitting curve of the dark I-V characteristics in the case of (b) Schottky and (c) Ohmic contact.

Full size image

Figure 7 (b) shows the comparison between the fit and the experimental data for the Schottky device. It can be seen that the fit of Eq. (1) is in good agreement with the data, so that the two barrier heights \(\:{\varphi\:}_{B}^{SC}=1.12\:eV\) \(\:{\varphi\:}_{B}^{MIS}=0.95\:eV\), the ideality factor η of 14, and the series resistance of 140 GΩ, were extracted as fitting parameters. It is interesting to note that the Schottky barrier obtained by fitting Eq. (1) is comparable to that calculated considering the work function of the Pt electrode (5.65 eV) and the electron affinity of the n-silicon (4.05 eV), which results in \(\:{\varphi\:}_{B}^{SC}=1.15\:eV\).

Regarding the Ohmic device, where the Schottky barrier is present only across the MIS junction, Eq. (1) becomes the equation of a single barrier. The comparison between the experimental data and the fitted curve for the Ohmic device, reported in Fig. 7 (c), reveals an excellent agreement, such that the barrier height \(\:{\varphi\:}_{B}^{MIS}\) and the quality factor η were extracted to be 0.9 eV and 1.9, respectively.

For the Ohmic device, the same barrier value \(\:{\varphi\:}_{B}^{MIS}\)=0.9 and the ideality factor equal to 3 was obtained by the Cheung-Cheung model based on the forward current analysis43,44 (the details are given in the Supplementary Information S4). The quality factor higher than 1, obtained by the two models, could be ascribed to the presence of the nitride layer between the graphene and the silicon and the possible presence of traps at the graphene\nitride interface45,46.

It could be noticed that the \(\:{\varphi\:}_{B}^{MIS}\) value obtained by the TE model is higher than that calculated by the FN approach for both devices. This result could be addressed to the different features of the mechanisms involved in the two models: in the FN model, the electrons tunnel through the partial width of the insulator barrier, while in the TE model, the charges pass over the insulator barrier (see Fig. 6(b) and (c)). Since the FN mechanism is bias dependent, unlike the TE model, and it is present only after the device switching V > VTh when the graphene Fermi level is shifted up16,47, the barrier that enables the tunnel through the insulator in the FN model is lower than that of the TE model.

Concerning the photoresponse of the Schottky device, illumination generates electron-hole pairs which, due to the built-in electric field across the MIS and Schottky junction, could head towards the electrodes without significant recombination. However, in the reverse bias state, the graphene Fermi level is higher than the silicon Fermi level, opening many states for photogenerated holes to be injected into, resulting in a high photocurrent16,19,48,49. Since the behavior of photocurrent versus voltage follows the same trend as the dark current (see Fig. 2(c) and Fig. 3(b)), the TE approach described by the Eq. (1) was used to fit the I-V curve of the Schottky device under illumination at 600 nm with a power of 0.2 µW. The model based on the double Schottky barrier is in good agreement with the data (see Supplementary Information S5), so the barrier heights were extracted to be \(\:{\varphi\:}_{B}^{SC}=0.95\:eV\) and \(\:{\varphi\:}_{B}^{MIS}=0.75\:eV\). This result confirms that illumination decreases the barrier due to the shift up of the Fermi energy level and the generation of photocarriers during the illumination process. In addition, it can be claimed that the presence of the Schottky contact on the back side not only guarantees rectification behavior but also reduces the recombination of photogenerated charges, directing them towards the external circuit and producing the high responsivity that characterizes the device.

In the Table 1, the performance comparison between the proposed SLGr/SiN/Si photodetector with similar structures is summarized. The results reported here demonstrate that the Schottky device represents the optimization of the previously realized device in which the same substrate structure was covered with a layer of graphene oxide (GO)10,11. The photoresponse has been improved over the spectral range from UV to NIR by replacing the GO with a graphene monolayer, which is gapless and avoids the pinning of the Fermi energy level thanks to its bias tunability. Moreover, the experimental results and data analysis demonstrate that the device based on the heterojunction SLGr/Si3N4/n-Si can be considered one of the most promising candidates for realizing a highly efficient photodetector in the UV-Vis-NIR range, combining the best features of 2D materials and conventional silicon technology.

Table 1 The performance comparison between the proposed SLGr/Si3N4/Si photodetector with similar structures is summarized.
Full size table

Conclusions

The study reports the successful development of a large-area graphene monolayer/n-Si heterojunction photodetector (MIS-like junction) with a Schottky junction on the back side. The realized device structure exhibits high photoresponse performance across the UV-Vis-NIR spectrum, surpassing that of a similar device with a rear ohmic contact. The Schottky-based device exhibits competitive characteristics, including a responsivity peak greater than unity at 685 nm, a specific detectivity of about 3 × 10¹² Jones, and a minimum noise equivalent power (NEP) of 1 × 10¹⁴ W/Hz¹/². These parameters are among the best reported in the literature for graphene/silicon-based detectors of comparable size (8 mm x 20 mm). An in-depth analysis of the current transport mechanisms, based on various conduction models, provides an estimation of the energy barrier height of the MIS junction to be approximately 0.9 eV in the dark and 0.75 eV under illumination.

The results demonstrate that the integration of a graphene monolayer within a metal-insulator-semiconductor (MIS)-like structure significantly improves the device’s photoresponse by reducing leakage currents, minimizing the recombination processes in the silicon, and facilitating the transport of photogenerated carriers. It is crucial to note that in the realized device, the graphene Fermi level can be tuned by combining the voltage across the graphene-based heterojunction with the presence of the Schottky junction on the back side. In summary, the results indicate that the graphene monolayer/n-Si heterojunction is a promising candidate for next-generation photodetectors, characterized by low cost, high performance, and CMOS compatibility.

Methods

An N-type Silicon wafer with a thickness of 300 μm and a resistivity in the range of 8–12 Ω∙cm was used as the substrate. The n-Si wafer was covered with a 60 nm thick Si3N4 layer deposited by plasma-enhanced chemical-vapor deposition (PECVD). Two circular Pt/Ti electrodes with a diameter of 1 mm were deposited on the top of silicon nitride. A 60 nm oxide layer (SiO2) was deposited only under the top electrodes, between the nitride and the silicon, to increase the isolation of the electrodes. In addition, two devices were fabricated depending on the backside structure of n-Si wafer: in one case, the Pt/Ti electrode is in direct contact with n-silicon (Schottky contact), and in the other case, a n + doped layer, obtained by ion implantation of phosphorus, has been interposed between the silicon and the Pt/Ti back electrode (Ohmic contact). In both devices, a single layer of graphene (SLGr) was transferred to the Si3N4 layer. Figure 1 shows the devices stacking sequence and the applied voltage bias. This configuration allows us to reveal the current drifting vertically across the heterojunction.

The graphene single layer (1 nm thick) was transferred to the Si3N4 layer, covering an active area of 40 mm2, using the following process. The SLGr, purchased from ACS Material (Pasadena, CA, USA), is sustained by a polymer substrate and covered with a 500 nm thick poly (methyl methacrylate) (PMMA) film. The thickness is about 1 nm as reported in the datasheet. The transfer of the SLGr on the Si-based substrates started with the immersion of the substrate in deionized (DI) water, where the PMMA–graphene film was released, floating on the water surface. Then, using the silicon-based substrate, the floating PMMA–graphene was picked up on it, then dehydrated, and finally heated on a hot plate. Eventually, the sample was steeped in acetone to dissolve the PMMA and patterned by O2 plasma etching after a photolithographic process to produce the desired layout.

The I-V characteristics in darkness and under illumination were carried out using continuum wave (CW) laser diodes ranging between 400 nm and 800 nm with a power ranging from 0.1 mW to 1mW. The surface of the device was irradiated using an optical fiber (with a spot diameter of 1 mm) mounted just in front of the sensitive area of the device. The I–V curves were measured using a voltage supply (Keithley Source Meter, model 2635) and a picoammeter (Keithley Dual-Channel, model 6482).