All-Graphene Photodetectors

We investigate the optoelectronic properties of novel graphene/FeCl3-intercalated few-layer graphene (FeCl3–FLG, dubbed graphexeter) heterostructures using photovoltage spectroscopy. We observe a prominent photovoltage signal generated at the graphene/FeCl3–FLG and graphene/Au interfaces, whereas the photovoltage at the FeCl3–FLG/Au interface is negligible. The sign of the photovoltage changes upon sweeping the chemical potential of the pristine graphene through the charge neutrality point, and we show that this is due to the photothermoelectric effect. Our results are a first step toward all-graphene-based photodetectors and photovoltaics.

H arvesting energy from photons is of tremendous societal importance since it contributes to reduce the carbon footprint. In common photovoltaic devices, the conversion from photons into electrical voltage is accomplished exploiting the in-built electric field at the interface of a p-and n-doped semiconductor to separate the photogenerated electronÀ hole pairs and originate a forward photovoltage. However, the intrinsic band gap of standard semiconductors restricts the photoresponsivity of these devices to specific light bandwidths. To harvest electricity over a wide range of the sun light spectrum, the multijunction design of stacked pÀn interfaces tuned to different bandwidths has been proposed. 1 Though these tandem solar cells display an improved light-harvesting efficiency, 2 they are typically brittle, heavy, and therefore difficult to implement in future flexible electronic devices.
Unlike conventional semiconductor materials, pristine single-and few-layer graphene (FLG) materials have no band gap. 3À6 The gapless energy dispersion allows electronÀ hole pairs to be generated over a broad light bandwidth from UV to THz. 7 Therefore, the implementation of all-graphene photovoltaics, that is, a device in which both the active area and electrodes are made of graphene materials, could be able to harvest energy over the whole sun light spectrum, while offering unique properties such as ultralight weight (i.e., graphene is just one atom thick), mechanical flexibility, and optical transparency. Therefore, understanding the optoelectronic properties of graphene-based heterostructures is the first step for exploiting the full potential of this carbon material in flexible and transparent photovoltaic devices.
Previous optoelectronic studies on graphene devices have shown that the photothermoelectric effect is at the origin of the measured photovoltage in graphene pÀn junctions and in single-bilayer interfaces. 8À10 On the other hand, the photovoltage measured at the grapheneÀmetal interface is due to a built-in electric field near the contact as a result of charge transfer from the metal contact to the graphene. 11À15 These graphene hybrid structures are at the core of a new generation of ultrafast photodetectors with a remarkable high bandwidth (500 GHz), zero source-drain bias (hence zero dark current) operation, and good internal quantum efficiency. 13 However, these devices still employ opaque metallic nanostructures which would introduce significant haze caused by light scattering 16,17 when used in smart windows and mirrors. The leap to all-graphene structures which have a high transparency would enable the development of a new generation of transparent photovoltaic devices which do not suffer from haze.
Here we study the photovoltage generated in novel all-graphene devices based on FeCl 3 -intercalated fewlayer graphene 18 (FeCl 3 ÀFLG, dubbed graphexeter) and pristine graphene. The FeCl 3 intercalation is known to dope graphene to record high charge carrier densities (up to ≈9 Â 10 14 cm À2 ), 18 and it drops the room temperature square resistance of graphene to just a few ohms, 18 making this material the best transparent conductor. At the interface between FeCl 3 ÀFLG/ graphene, we observe a dominant photovoltage comparable to the signal measured at the graphene/Au interface. We observe a sign reversal of the photovoltage upon sweeping the chemical potential of the pristine FLG through the charge neutrality point, and we show that this is due to the photothermoelectric effect. Our results demonstrate that FeCl 3 ÀFLG can replace expensive and opaque metals in photovoltaic architectures, making these structures mechanically flexible and transparent.
The device is fabricated as follows: pristine few-layer graphene is first deposited by mechanical exfoliation onto heavily doped Si/SiO 2 substrate. Raman spectroscopy and optical contrast are used to determine the number of graphene layers as well as their stacking order (see Table 1 and Supporting Information for a summary of the key device parameters). The intercalation with ferric chloride is performed at 360°C at 2 Â 10 À4 Torr for 7.5 h following the procedure described by Khrapach et al. 18 (also see Methods). During this process, ferric chloride molecules penetrate between the layers of FLG and heavily p-dope it to record high levels of ≈9 Â 10 14 cm À2 . Subsequently, a pristine FLG flake is transferred over the FeCl 3 ÀFLG flake following the methods described in refs 19 and 20. Independent multiple electric contacts made by Cr/Au (5 nm/70 nm) to the bottom FeCl 3 ÀFLG and top FLG allow the characterization of these novel heterostructures (see Figures 1 and 2a). Figure 1 shows the schematic of the final device along with the measurement setup. The FeCl 3 ÀFLG flake is kept on the ground, while a small dc bias of 0.1 mV is applied to the pristine FLG flake. The graphene photodetector is then illuminated by a 532 nm HeNe laser focused by using a 100Â objective to 1.5 μm spot size at a power of 8.2 μW. The beam is chopped at 370 Hz, and the chopper is used as reference to a lock-in amplifier which measures the photovoltage. The heavily doped Si substrate acts as a   ARTICLE global back-gate which we use to tune the chemical potential of graphene, whereas the resistivity of the FeCl 3 ÀFLG flake is unaffected by the typical values of used gate voltage due to the high doping level. 18 The devices are mounted on a scanning stage which allows one to map the photoresponse of these graphenebased heterointerfaces in the xÀy directions with a spatial resolution of 1 μm.

RESULTS AND DISCUSSION
To probe the homogeneity of the intercalation, we employ Raman spectroscopy (see Methods). Figure 2a shows a map of the Raman G-band for the nonintercalated and intercalated parts for a typical device. It is apparent that the pristine FLG shows the wellknown strong Raman intensity at 1580 cm À1 corresponding to the G-band, whereas a strong Raman intensity at 1610 cm À1 is present over the whole area of the intercalated FLG, demonstrating the uniformity of the intercalation process. The upshift of the G-band to 1612 and 1625 cm À1 has been previously studied and attributed to charge transfer from FeCl 3 to graphene. 18,21 More specifically, the shift of the G-band to 1612 cm À1 is a signature of a graphene sheet with only one adjacent FeCl 3 layer, whereas the shift to 1625 cm À1 characterizes a graphene sheet sandwiched between two FeCl 3 layers. Figure 2b shows the Raman G-band for the pristine (blue) and intercalated (red) FLG of the device. The pristine FLG is lightly doped as indicated by the Raman G-band appearing at 1583 cm À1 . 22,23 On the other hand, the Raman spectrum of the intercalated FLG shows three shifts of the G-band to 1587, 1608, and 1610 cm À1 . These observations suggest that the peak at 1610 cm À1 originates from two graphene layers sandwiching one layer of FeCl 3 , 18,21 whereas the other peaks are due to doped pristine graphene layers. 18 Since the intercalated flakes which we chose for these experiments are trilayers, the structure can be understood as a layer of FeCl 3 sandwiched between a graphene monolayer and a graphene bilayer, as schematically shown in the inset of Figure 2b.
To further characterize the devices, we study the electrical transport properties of the independently contacted pristine FLG and FeCl 3 ÀFLG flakes. Electrical measurements are preformed in constant current using an excitation current of 100 nA in four-terminal configuration to avoid the contact resistance at the interface with metals. 24 A summary of the backgate and temperature dependence of the resistance of a typical device is presented in Figure 2c,d. The FeCl 3 ÀFLG shows no gate control of the resistivity (red curve in Figure 2c), which is typical of heavily doped graphene. On the other hand, the pristine FLG (see blue curve in Figure 2c) exhibits the expected large modulation of resistance as a function of gate voltage and a maximum resistance at V g = 40 V for this specific device. This indicates the presence of residual p-doping probably caused by FeCl 3 molecules present on the surface of the underlaying FeCl 3 ÀFLG. Consistently, we observe that the FeCl 3 ÀFLG has a room temperature resistivity of ≈11 Ω which decreases upon lowering the temperature down to 9 Ω at 4.2 K (see Figure 2d). The observed metallic behavior of the resistivity is consistent with the heavy p-doping of the system induced by the intercalation with FeCl 3 , and it is contrasted by the typical temperature dependence of the pristine FLG, which shows increasing resistivity upon lowering temperature 25 (see Figure 2d).
The optoelectronic properties of these graphenebased hybrid structures are characterized by measuring the photovoltage generated across the pristine FLG/FeCl 3 ÀFLG interface while rastering the laser spot over the active device area. Figure 3a shows the photovoltage generated in device D1 as a function of position of the laser beam (see Supporting Information for a similar characterization of another device). It is apparent that there is a strong photovoltage at the Au/ FLG (blue) and FLG/FeCl 3 ÀFLG (red) interfaces, while the photovoltage at the FeCl 3 ÀFLG/Au is nearly 0. To understand the origin of the generated photovoltage, we fixed the position of the laser beam on a specific location of the interfaces, and by changing the back-gate voltage, we modulated the chemical potential from holes to electrons in the pristine FLG. Figure 3b,c shows the gate dependence of the resistance for the different interfaces found in the device as indicated in the graph. In particular, for the pristine flake, the charge neutrality point (CNP) occurs at 20 V and the crossover from hole transport to electron transport can be studied (see Figure 3b,d).
A comparison of the gate dependence (in the range of À50 V < V g < 70 V) of the photovoltage for all interfaces shows striking differences in the measured signal depending on the interface that is measured (see Figure 3d,e). More specifically, experimentally, we find no detectable photovoltage generated at the Au/FeCl 3 ÀFLG (black). This is in contrast to the photovoltage generated at the Au/FLG (blue) and FLG/ FeCl 3 ÀFLG (red) which is nonzero (up to (30 mV/W), it is nonmonotonous, it switches sign when the gate voltage drives the Fermi energy accross the charge neutrality point, and it decreases monotonously for very high doping levels. Furthermore, we observe that the photovoltage generated at the Au/FLG is of comparable magnitude to that measured at the FLG/ FeCl 3 ÀFLG interface but it has opposite sign, that is, negative in the hole side and positive in the electron side. Finally, we note that the photovoltage generated at graphene/FeCl 3 ÀFLG is equivalent or larger than what has been previously reported in doubly gated graphene pÀn junctions.
The fact that (1) the photovoltage generated at the Au/FLG (blue) and FLG/FeCl 3 ÀFLG (red) is nonzero, (2) it is nonmonotonous, (3) it switches sign when the gate voltage drives the Fermi energy across the charge ARTICLE neutrality point, and (4) it decreases monotonously for high doping levels suggests that photothermoelectric effects are at the origin of the observed signal. Similar observations have been previously reported in graphene single-bilayer junctions 8 as well as doubly gated graphene pÀn junctions, 9,10 demonstrating the occurrence of photothermoelectric effects later supported by theoretical studies. 26 In the photothermoelectric effect, after the electrons are photoexcited from the valence band to the conduction band, they form a hot Fermion distribution by relaxing back to the Fermi level. When there is a difference in the density of states of the materials forming the photoactive interface, the carriers diffuse from one side to the other of the interface, leading to a photovoltage. 8,9,26,27 If charge transfer was responsible for the measured photovoltage, the signal should have increased upon increasing the carrier concentrations as the band bending increases. 26,27 This is in contrast with the observed decrease of the photovoltage for high doping of the non-intercalated flake, which is instead expected for photothermoelectric effects. Finally, in our devices, we do not expect the contacts to contribute with significant photovoltaic effects since our interface material is chromium, which is known to induce a very small band bending in graphene. 27À29 Having established that photothermoelectric effect is at the origin of the observed photovoltage, the measued zero photovoltage at the FeCl 3 ÀFLG/Au interface implies that the Seebeck coefficients of FeCl 3 ÀFLG are similar to that of Au. This observation, together with the fact that the magnitude of the photovoltage measured at the FLG/Au and FLG/FeCl 3 ÀFLG is comparable, demonstrates that FeCl 3 ÀFLG is a good replacement for metals or local gates in future graphene photodetectors.
The photovoltage generated by the photothermoelectric effect is V pv = (S 2 À S1)ΔT, where S i is the Seebeck coefficient of the different materials and ΔT is the temperature difference. The Mott relation gives 30À32 For device D1, the top layer graphene is ABA trilayer graphene (see Supporting Information), and we

ARTICLE
approximate the E f (n) dependence to be that of bilayer graphene where 33 Here, γ 1 is the interlayer coupling strength, which we take to be 0.4 eV. 34,35 Figure 4a shows the dependence of the electrical conductance (G) as a function of the charge density (n), where n is extracted from V g using the plane plate capacitor model. Figure 4b shows the calculated Seebeck coefficient using the measured G(n) and eqs 1 and 2. The measured photovoltage has a similar charge density dependence to the Seebeck coefficient, and both signals cross over from positive to negative at the charge neutrality point (see Figure 4c). This has to be expected when the photothermoelectric effects dominate the measured photovoltage. 8À10 In these devices, only the Seebeck coefficient of the ABA trilayer flake contributes significantly to the photovoltage since the Seebeck coefficient of the FeCl 3 ÀFLG is 0 as there is no gate modulation of the resistivity due to the large density of states ; that is, dG/dn ≈ 0 for our available gate voltage range. Furthermore, Figure 4b,c shows that the Seebeck coefficient and the measured photovoltage are not exactly proportional. This discrepancy can be attributed to the local differences in the magnitude of the Seebeck coefficient induced by inhomogeneous doping of the ABA trilayer graphene flake since the photovoltage is a probe of the local density of states. 8

CONCLUSIONS
In summary, we show that the FeCl 3 ÀFLG can be used as a replacement for metal contacts in graphene photodetectors. The high degree of doping in FeCl 3 ÀFLG makes the resistivity of this material insensitive to a global gate voltage acting on the pristine graphene; therefore, FeCl 3 ÀFLG is an ideal conductive interconnect material. Our experiments demonstrate that FeCl 3 ÀFLG can replace metals in a new generation of all-graphene-based photodetectors. In particular, we demonstrate a maximum photovoltage of ≈0.1 V/W at the FLG/FeCl 3 ÀFLG interface. We attribute the measured photovoltage to the photothermoelectric effect with a maximum Seebeck coefficient of 20 μV/K for ABA trilayer graphene.

METHODS
Intercalation with FeCl 3 . The intercalation process with FeCl 3 is performed by placing both the anhydrous FeCl 3 powder and the substrates with pristine few-layer graphene in vacuum (∼10À4 mbar) and at 310 and 350°C for 7.5 h. A heating rate of 10°C/min is used during the warming and cooling of the two zones.
Raman Spectroscopy. Raman spectra are collected in ambient air and at room temperature with a Renishaw spectrometer. An excitation laser with a wavelength of 532 nm, focused to a spot size of 1.5 μm diameter and a ∼100Â objective lens, is used. To avoid sample damage or laser-induced heating, the incident power is kept at 5 mW.