Hybrid graphene tunneling photoconductor with interface engineering towards fast photoresponse and high responsivity

Hybrid graphene photoconductor/phototransistor has achieved giant photoresponsivity, but its response speed dramatically degrades as the expense due to the long lifetime of trapped interfacial carriers. In this work, by intercalating a large-area atomically thin MoS2 film into a hybrid graphene photoconductor, we have developed a prototype tunneling photoconductor, which exhibits a record-fast response (rising time ~17 ns) and a high responsivity (~$3\times10^4$ A/W at 635 nm and 16.8 nW illumination) across the broad spectral range. We demonstrate that the photo-excited carriers generated in silicon are transferred into graphene through a tunneling process rather than carrier drift. The atomically thin MoS2 film not only serves as tunneling layer but also passivates surface states, which in combination delivers a superior response speed (~3 order of magnitude improved than a device without MoS2 layer), while the responsivity remains high. This intriguing tunneling photoconductor integrates both fast response and high responsivity and thus has significant potential in practical applications of optoelectronic devices.


Introduction
Photodetectors with high speed and responsivity are of crucial importance for optoelectronic applications. Benefiting from their large surface area and unique electronic properties, twodimensional (2D) materials are promising building blocks in optoelectronic devices. [1,2] The gapless semi-metallic band structure of graphene gives rise to its high mobility, extremely broad optical absorption and ultrafast carrier dynamics, [3][4][5][6] making this emerging 2D material particularly attractive in photodetection applications. [7,8] However, the responsivity of phototransistor comprised of pristine graphene is limited to the level of mA/W due to the low optical absorption rate and fast carrier recombination in graphene. [8,9] Recently researchers have proposed an effective method to improve the responsivity of graphene-based photodetector by adding a light absorption material, such as PbS quantum dots, [10] Bi2Te3, [11] perovskites, [12,13] silicon (Si), [14,15] carbon nanotubes, [16] atomically thin MoS2, [17][18] and organic crystal, [19] as active layer and mainly remaining the graphene as the channel transport layer. This hybrid graphene photoconductor/phototransistor introduces a vertical built-in field at the interface for exciton dissociation. The photo-excited carriers in the active layer can be efficiently transferred into graphene, of which the electrical conductance is highly sensitive to external carrier injection. [20] Although this device configuration brings great enhancement of responsivity, the response speed of graphene-based photoconductor drastically degrades as the cost. The interface trap states, together with the drift time of photo-excited carriers in the depletion region, increase the response time into the range from several microseconds to several seconds. [10][11][12][13][14][15][16][17][18][19] This trade-off between response speed and responsivity makes it a great challenge to achieve fast and high-gain photoresponse simultaneously in graphene-based photodetectors, severely limiting their potential application for photodetection.
Here we propose a high-performance atomically thin hybrid graphene tunneling photoconductor constructed with light absorption and interface passivation layers. This hybrid graphene photoconductor exhibits a response time of ~17 ns, which is improved by ~3 orders of magnitude compared with the device without MoS2 intercalation and is at present the shortest one among those of hybrid graphene photoconductors/phototransistors, along with a high responsivity across a broadband spectral range (~3×10 4 A/W at 635 nm illumination with 16.8 nW power). In our device, the Si serves as an optically active layer and MoS2 film acts as a passivation layer for reducing surface states, suppressing carrier recombination at the interface, and offering an ultra-thin layer for the carrier tunneling. Carrier transfer process driven by ultrafast quantum tunneling effect rather than by the carrier drift in the depletion region enables a superior response speed, while the responsivity remains high as a major merit of hybrid graphene photoconductor/phototransistor. The novel device designs and desirable results not only provide a unique structure for better understanding the photocarrier transfer and transport processes in 2D layered nanomaterials but also pave the way for the development of state-ofthe-art graphene-based photodetector with both high speed and responsivity for practical applications. (n-Si, carrier density ~10 15 cm -3 ) wafer was used as the starting substrate. A 100 nm SiO2 thin layer was grown on the Si wafer and a 105 μm ×115 μm Si window was defined by CF4 etching.

Device Structure
Afterward, monolayer MoS2 film grown by chemical vapor deposition (CVD) was transferred onto the Si groove region of the substrate, and a subsequent pattern process was carried out to enable the remaining MoS2 cover the Si window. After monolayer graphene film was transferred onto the MoS2/Si heterostructure, a pair of Au source/drain electrodes was thermally deposited atop graphene to form a channel with 125 μm in length. Figure 1c displays the Raman spectra taken in the vertical heterostructure area of the device (violet dashed rectangle in Figure   1b). Characteristic Raman signals of both MoS2 and graphene were clearly detected. The feature peaks of MoS2 at 384.6 cm -1 (E 1 2g) and 403.1 cm -1 (A1g) with an 18.5 cm -1 difference indicate the monolayer nature of the MoS2 film. [29] The high quality of monolayer graphene was also confirmed by the absence of D band and the high intensity ratio of 2D and G peaks. [30] Figure   1d shows a typical optical image of large-scale continuous monolayer MoS2 film before transfer.
Photoluminescence (PL) measurement of the MoS2 film (upper left inset of Figure 1d) reveals a direct optical bandgap of 1.82 eV, which reconfirms its monolayer feature. [31] High-resolution transmission electron microscopy (HRTEM) image in the upper right inset of Figure 1d presents a highly crystalline structure of monolayer MoS2.   Figure 2 provides the photoresponse characteristics of the hybrid graphene photoconductor.

Photoresponse Performance
Typical illumination power-dependent response at a wavelength of 635 nm has been measured, as shown in Figure 2a. The device shows negative photoresponses because photo-excited holes are injected into n-type graphene while electrons are trapped in Si, lowering graphene's quasi-Fermi level through photo-gating effects. [10,14,32] Both MoS2 and n-Si can bring n-dopants to graphene (see Supporting Information S1), [14,18,33] thus the Fermi level of graphene prior to being illuminated is above its Dirac point. The work function of graphene in the channel area is measured to be ~4.4 eV with a Kelvin probe force microscopy (KPFM) system ( Figure S1b).
The photocurrent ph , which is defined as the absolute difference between light current light and dark current dark ( ph = | light − dark |), shows a monotonic increase as the light power increases. The responsivity, together with the photocurrent as functions of illumination power at d = 5 V are given in Figure 2b. The photoconductor shows a considerably high responsivity of ~3×10 4 A W -1 at the light power of 16.8 nW, which far exceeds that in pure graphene device, and has the same order as that in graphene/Si photoconductor without MoS2 intercalation (see Supporting Information S2). As the responsivity increases with decreasing light power, a higher responsivity is expected in ultra-weak illumination. The corresponding photoconductive gain is calculated to be ~5. where τ lifetime is the lifetime of injected carriers from Si and τ tr is the carrier transit time across the channel. [34] The trapped photo-excited electrons in Si enable the excess hole conservation (with long τ lifetime ) in graphene by capacitive coupling, leading to the carrier reinvestment in graphene channel and the ultrahigh gain in the photodetector. [10,14,35] Wavelength-dependent photoresponses at 5 V bias were measured via quartz tungsten halogen lamp filtered with a monochromator. Normalized photocurrent versus light wavelength plotted in Figure 2c reveals a broad spectral response with considerable responsivity from visible light to the near-infrared region (~1120 nm). Relatively higher photoresponse than pure Si photodetector from violet to green is assigned to additional absorption by monolayer MoS2. Normalized photocurrent versus illumination wavelength.

Transient Performance
Temporal response of the hybrid graphene photoconductor was conducted under electrically modulated 635 nm light pulse having illumination power of 3.35 μW. Figure 3a shows the transient characteristics of the device at repeated frequency of 2 Hz, which demonstrates a highly repeatable and stable switching characteristics between light-off and light-on states.
When the repeated frequency of light input increases to 100 kHz, the photoconductor can maintain its promising switching performance (Figure 3b). The fast rising edge followed by quick falling tail in the light-switching-on front results from a competition between hole drifting and electron diffusion processes in the depletion region. [15] The spikes in light switching points are due to the limitation of the pulse generator.
In order to further investigate the response speed of the device, which reveals the capacity to follow high-frequency optical signals, the detailed information of the rising and falling edges in a single switching cycle is provided in Figures 3c and 3d can be identified. Note that some oscillation signals with a period of ~28 ns occur at the rising and falling edges; it is suggested to be the intrinsic RC oscillation of the operating circuit. By measuring the resistance and capacitance of the operating circuit (graphene channel) (2.3×10 2 Ω and 2.1×10 −11 F, respectively), we obtain the calculated RC time constant RC = 2 ≈ 30 ns, which is self-consistent with the directly measured value from the oscilloscope.
The phenomenal response speed of the tunneling graphene-based photoconductor far exceeds those in previous reports on hybrid graphene photoconductors/phototransistors, and is even close to the limitation of intrinsic circuit oscillation. Furthermore, in stark contrast to commercial Si photoconductors/phototransistors of which response times are around several tens of microseconds (e.g., OP602TX Si phototransistor from OPTECK with rise~2 0 μs), [36] this hybrid graphene photoconductor overcomes the limitation of pure Si device by enhancing the response speed of about 3 orders. This results from the photo-excited carriers in our device transiting across the graphene channel, which is a much faster process than the carriers transiting across the Si channel in pure Si photoconductors. Such fast photoresponse makes this device highly promising in high-frequency optical sensing applications such as military warning, highspeed imaging and chemical sensing.

Photocarrier Transport Process
We have proposed a synergistic mechanism to explain the extraordinary performance of the hybrid graphene photoconductor with MoS2 interface layer, which will be demonstrated in the following. Figure 4a illustrates the band diagram of the device. A Schottky barrier is formed at the interface between graphene and Si. transport. [25,37] The built-in field separates the EHPs and consequently the holes are transferred into graphene, leading to the conductance reduction of graphene. It is estimated that the quasi Fermi level of graphene is pulled down by 9.2 meV due to the photo-excited holes injection under 3.35 μW illumination and the built-in barrier height bi prior to being illuminated is ~0.15 eV (see Supporting Information S4). This Fermi level lowing caused by illumination is also detected by the Raman G peak redshift in n-type graphene with increasing laser power ( Figure S4b). Here the MoS2 film plays several crucial roles in device operating. First, MoS2 passivates the interface of the Schottky junction to reduce trap states and to further suppress carrier recombination. Meanwhile the ultra-flat MoS2 underneath graphene allows the intrinsic electronic properties to be retained in graphene. [38] Second, the relatively wide electronic bandgap of monolayer MoS2 (~2.4 eV) results in a straddling gap at MoS2/Si interface, namely, type-I heterostructure. [39,40] In this band alignment, photo-excited holes are accumulated at the interface triangular well (green circles in Figure 4a), while electrons are blocked from transiting towards graphene. The valence band offset at MoS2/Si (∆ V ) around 1.1 eV [39] provides a barrier to prohibit the thermionic emission for the photo-excited holes (thermal energy at 300 K is ~26 meV), thus suppressing the thermal noise. Third, the atomic thickness of MoS2 enables the carrier wave function in Si to interact with that in graphene, causing the carrier quantum tunneling through MoS2 layer. The band bending at MoS2/Si interface (Figure 4a) generates a strong built-in field that further enhances the hole tunneling.

Photocarrier Tunneling
The photo-excited carriers (holes) in the device are injected into graphene via an ultra-fast tunneling process. The response time rise is affected by the drift time for photo-excited carriers to travel from depletion region to the Schottky junction interface ( drift ), the average time for carriers to transit through interface trap states ( trap ), the average time for carriers to tunnel through MoS2 film ( tunnel ) and carrier transit time in graphene channel ( tr ). tr is estimated to be tr = µ graphene d / ≈ 10 ns, where is the channel length (125 μm) and µ graphene is carrier mobility in graphene which has been derived to be ~3.1×10 3 cm 2 V −1 s −1 (see Supporting Information S4). trap is greatly lessen with MoS2 passivation, as discussed above.
drift is also reduced when MoS2 layer is used since MoS2 shares the depletion region of the Schottky junction and makes the length of remaining depletion region in Si shorter. tunnel is highly sensitive to the tunneling barrier (MoS2 film) thickness , which is written as below based on the Wentzel-Kramers-Brillouin approximation: [34,41] tunnel ∝ 1/ tunnel ∝ exp [2 √ ( where tunnel is the carrier tunneling probability, p * is the effective mass of the accumulated hole in Si, and tunnel is the average barrier height for tunneling. The carrier tunneling phenomenon originates from the wave nature of carriers in quantum mechanics, and the tunneling time tunnel is not determined by the classic transit time process (i.e., = / , where is the carrier velocity). [34] Tunneling time can be very short when the tunneling barrier is ultrathin (<2 nm), typically in the timescale of subnanosecond. [41] The high speed response in the photoconductor benefits dominantly from this ultra-fast tunneling process when the MoS2 layer is thin enough. Additionally, graphene based Schottky junction enables the tunneling effect to occur in relatively small built-in barrier while conventional TFETs require a gate voltage for band-to-band tunneling. [28] To certify the tunneling mechanism described above, we investigate the thickness effect of the MoS2 passivation layer on the photoconductor performance. Figure 4b depicts the energyband diagram of the hybrid graphene photoconductor with multilayer MoS2 interface layer. As the tunneling barrier gets thicker, the carrier tunneling probability tunnel drastically reduces (Eq. (1)). Correspondingly, not only the tunneling time tunnel which is inversely proportional to tunnel becomes much longer, but also the total population of photo-excited carriers which can tunnel into graphene decreases. We have fabricated additional devices with multilayer MoS2 film (~4 layers, confirmed by Raman spectrum, Figure S5a), with all the other factors unchanged. Temporal response measurements show that the photoconductor with thicker interface layer presents a response time of ~500 ns (Figure 4c), revealing that larger tunneling barrier thickness causes a longer tunnel (Eq. (1)). Moreover, the photoresponsivity in the photoconductor with multilayer MoS2 is much smaller than that of device with monolayer MoS2 (Figure 4d), which well agrees with the above analysis on the injected carrier population decrease (note that the increased light absorption by MoS2 is limited). [42] Tunneling Accumulation To further understand the photocarrier tunneling process, we investigate the temperaturedependent carrier transport in the device operating in diode mode (i.e., the source and drain electrodes are on graphene and Si, respectively).
where A is the junction area, A * is the effective Richardson constant, is the absolute temperature, is the Boltzmann constant, and is the ideality factor. The reverse saturation current at diode bias bias = −2 V ( dark s ) as a function of temperature is plotted in red circles in Figure 4f (Arrhenius plot). A linear l g( dark s )-1/ dependence is found, demonstrating the carrier transport in the dark is a thermally activated process with the express dark ∝ exp (− a ), where a is the activation energy. [34] When illuminating the diode at zero bias, the current is fully contributed by photo-excited carriers ( ph = light at bias = 0), and the holes tunnel through MoS2 layer into the graphene rather than being thermally activated (right inset in Figure 4e). As illustrated in Figure 4f, ph at zero bias is temperature insensitive. The shortcircuit hole tunnel photocurrent is given by: [37] ph = 4 p * ( ) 2 where V is the effective density of states in the valence band of Si, and s is the concentration of holes at the interface. This expression indicates a non-thermally-activated process of the photo-excited carrier since Δ Fp promotes the carrier tunneling rather than setting up an energy barrier. From the Richardson plot of ph (l g( ph / 2 )-1/ plot) as illustrated in Figure S6, one can identify that ph / 2 increases with decreasing temperature, which is in good coincidence with Eq. (3). The above analysis of photo-excited carrier transport at zero bias in the graphene/Si diode with MoS2 layer can be fully applied to the device operating in photoconductor mode since there is no external vertical bias across the photoconductor (Figure   1a).

Interface Passivation
In previous studies of graphene photoconductor/phototransistor hybridized with a semiconducting active layer, the photo-excited carriers were transferred to graphene driven by the built-in field. During this process, the carriers can be trapped by interface states, thus resulting in long carrier transfer time. Here, by inserting an atomically thin MoS2 layer into hybrid graphene photoconductor, the interface traps are passivated and thus the photo-excited carriers can be efficiently transferred into graphene through the ultra-fast tunneling process. We further investigate the MoS2 passivation effect by the noise spectra of the devices. It is widely accepted that the noise in metal-oxide-semiconductor transistors can be explained by the trap charge fluctuation mechanism. [43] The interface trap states are considered to be one of the dominant contributions to the low-frequency 1/f noise in monolayer graphene devices. [44][45][46] It is observed that the graphene-based photoconductor with MoS2 interface layer performs more than one order of magnitude lower noise power density ( / 2 ) when compared with the device without MoS2 layer (Figure 5a). The average noise amplitude Anoise, which is defined as , can quantitatively probe the level of 1/f noise for a given frequency range. [44] For devices without and with MoS2 passivation, Anoise is calculated to be 1.5×10 -8 and 2.4×10 -10 , respectively at low frequencies . This large noise reduction indicates that the interface traps are greatly removed by MoS2 passivation, demonstrating that MoS2 serves as perfect substrate for graphene. [38] By designing the novel graphene-based photoconductor with interface engineering, we have approached the highest response speed among hybrid graphene photoconductors/phototransistors. For detailed comparison, we have summarized the figures of merit (response time and responsivity) for some previously reported hybrid graphene photoconductors/phototransistors and our device, as shown in Figure 5b. [10][11][12][13][14][15][16][17][18][19]

Flexible Device
Furthermore, this hybrid graphene photoconductor does not require electrical gating to operate, which makes it possible to obtain flexible device if Si thin film (thickness ~200 nm) exfoliated from silicon-on-insulator is adopted (see Supporting Information S7). Here we fabricated a flexible device on a polyimide substrate, as manifested in Figure S7a. This flexible device shows the high reliability of photoresponse under 635 nm illumination after mechanical bending for 100 times (Figures S7b-S7c). The responsivity of the flexible photoconductor under 168 nW light power is ~1×10 3 A/W, and the response time rise is around 0.2 s. The flexible device exhibits competitive performance when comparing to previously reported flexible photodetector based on graphene [47] and other 2D materials. [48] Due to the large amount of traps in polyimide and Si thin film, the flexible device cannot response as fast as the rigid one does at present.

Conclusion
In summary, we developed a novel interface engineered hybrid graphene photoconductor with a record-fast response within 17 ns and a high photoresponsivity across a wide spectral band. The atomically thin MoS2 film in the device not only passivates interface traps but also acts as a barrier for photo-excited carrier tunneling into graphene, which accounts for the remarkable response speed. We have investigated the tunneling mechanism in detail and identified the importance of incorporating tunneling barrier in atomic thickness. Temperatureinsensitive short-circuit photocurrent in graphene/Si diode with MoS2 interface layer provides strong evidence for the photo-excited carrier tunneling in the device. Also, the interface passivation by MoS2 is demonstrated through noise analysis. This demonstrated graphene-based photoconductor with interface engineering opens a pathway towards ultra-fast, high responsivity and broadband photodetection with the large photo-active area, therefore providing a myriad of opportunities for future study and practical applications of 2D materials.

Methods
CVD growth of MoS2: Monolayer/multilayer MoS2 films were fabricated using a home-built CVD system. [49] A quartz boat holding ammonium molybdate powder (6 mg) for producing uniform MoO3 during growing was put in the center of a 2-inch CVD quartz tube. SiO2/Si substrates treated with acetone and O2 plasma were placed next to the boat. Another quartz boat holding precursor of S powder of 0.7 g was placed 30 cm away from ammonium molybdate.
The reaction furnace was heated to 800 °C at the rate of 15 °C min -1 and S powder was heated to 180 °C when the reaction zone reached the temperature of 650 °C with a 50 sccm Ar flow at the pressure of ~100 Pa. After ~30 min MoS2 growing, the furnace was cooled down to room temperature and MoS2 films with high quality were deposited on the substrates.
Characterization: A Renishaw Spectroscope with 514.5 nm laser excitation was applied to perform Raman and PL measurements. The size of laser spot was ~1 μm. HRTEM measurements were conducted via FEI Technai F20 system. KPFM measurements were performed using a Bruker Dimension Icon system.

S4. Fermi level shifting of graphene induced by photo-excited carrier injection and the built-in barrier height
Because of the low density of states near the Dirac point in graphene, its quasi-Fermi level F ′ is easily effected by excess carrier injection according to the equation S4 where F ≈ 10 6 m/s is the Fermi velocity of the massless Dirac fermion in graphene, F ′ is the Fermi wavevector magnitude, is the carrier density in graphene at equilibrium and ∆ is the excess carrier density due to illumination. in Figure S4a, showing that as the light power increases to 3.35 μW, F ′ is pulled down from 161.8 meV to 152.6 meV. Raman G peak shift is known to probe the doping level in graphene. S5 We have taken Raman measurements of graphene G band in the device with various Raman laser powers, as shown in Figure S4b. As the laser power increases from 10% to 100%, the G peak shows a redshift from 1586.2 cm -1 to 1582.3 cm -1 , indicating the graphene is getting less doped.
The surface potential of n-Si adopted in the device is around 4.27 eV, therefore the initial built-in barrier height bi of the device can be estimated to be 0.13 eV. Capacitance-voltage (C-V) measurement of the barrier without illumination, as is shown in Figure S4c, indicates the barrier height is ~0.15 eV, which is consistent with the estimation from KPFM measurements.

S7. Flexible hybrid graphene photoconductor
The fabrication process of the flexible device is very similar to that of rigid device. The major difference is the usage of Si thin film exfoliated from SOI rather than bulk Si. Firstly, SOI is put into HF solution (HF:H2O=1:4) for etching SiO2. Then the Si thin film (thickness ~200 nm) is transferred onto the polyimide substrate and patterned to rectangle by CF4 etching. The following steps, e.g., transfer and patterning of MoS2 film, photolithography and electrode deposition are the same with those of rigid device, as described in our main manuscript. Figure   S7 summarizes the schematic diagram and the photoresponse performance of the flexible device.