NANO LETTERS
Imaging of Photocurrent Generation and Collection in Single-Layer Graphene
2009 Vol. 9, No. 5 1742-1746
Jiwoong Park* Department of Chemistry and Chemical Biology, Cornell UniVersity, Ithaca, New York 14853
Y. H. Ahn DiVision of Energy Systems Research, Ajou UniVersity, Suwon 443-749, Korea
Carlos Ruiz-Vargas School of Applied and Engineering Physics, Cornell UniVersity, Ithaca, New York 14853 Received September 27, 2008; Revised Manuscript Received February 23, 2009
ABSTRACT Unlike in linear nanostructures, photocurrent generated in single-layer graphene (SLG) is expected to display two-dimensional characteristics. This allows the investigation of carrier dynamics, in relation to several spatially varying factors (such as the location of photocurrent generation and collection) and the overall electron band configuration of the SLG. In this letter, we use scanning photocurrent microscopy to investigate the spatial mapping of photocurrent generation and collection in SLG in a multielectrode geometry. A strong electric field near metal-graphene contacts leads to efficient photocurrent generation, resulting in >30% efficiency for electron-hole separation. The polarity and magnitude of contact photocurrent are used to study the band alignment and graphene electrical potential near contacts, from which it is shown that there exist large-scale spatial variations in graphene electric potential. Our measurements with a multielectrode device configuration reveal that photocurrent is distributed with a clear directional dependence among different collector electrodes. In the same measurement scheme, we also determine the majority carrier in graphene under different gate conditions by imaging the thermocurrent generated by laser-induced heating of electrodes.
Graphene, which is a single-layer atomic carbon crystal, has recently attracted a great level of excitement and attention, thanks to its unique combination of electrical, mechanical and optical properties.1-4 In addition to having great potential as a next-generation electronic material, graphene is also suggested to be an excellent candidate material for novel photovoltaic devices.5-7 Despite the absence of a band gap, several layers of graphene can form an interesting light harvesting layer, because of its broadband light absorption and large carrier mobilities. Single-layer graphene (SLG), on the other hand, can also be used to produce transparent electrodes with great broadband transparency.5,8 Several key results have been reported, regarding the other optical properties of graphene, including gate-tunable optical transitions in infrared reflectivity9 and visible absorption spectroscopy.10 More recently, scanning photocurrent microscopy (SPCM) has been applied by Lee et al.,11 to study the photoelectric properties of SLG devices in conjunction with the local potential variation near metal contacts and sheet edges. Although some interesting data have been * Author to whom correspondence should be addressed. E-mail:
[email protected]. 10.1021/nl8029493 CCC: $40.75 Published on Web 03/27/2009
2009 American Chemical Society
presented in this recent report, the physical implications regarding the generation and collection of photoexcited carriers in a multielectrode device geometry have not been discussed in detail. In this paper, we report spatially resolved scanning photocurrent (PC) and thermoelectric (TE) imaging in a SLG-based device with multiple electrodes. In both imaging techniques, a focused laser beam is scanned over a device made with SLG while the electrical signal from the device is measured as a function of the position of the laser spot to construct a spatial image with a diffraction-limited resolution. Our setup also measures the reflection signal simultaneously to accurately determine the absolute position of the laser beam. This SPCM technique was previously used to study devices based on individual carbon nanotubes12-15 and semiconducting nanowires,16-18 where the PC is confined to an essentially one-dimensional current pathway. Graphene, on the other hand, provides a two-dimensional (2D) conducting sheet and allows many different spatial routes for PC generation, transport, and collection. Therefore, SPCM provides an ideal tool for investigating photoelectric behaviors of graphene. In our experiment, a focused laser beam
Figure 1. Photocurrent imaging of a graphene device. (a) Optical image of the graphene device (the channel defined by electrodes is 12 µm long and 2.5 µm wide), scale bar ) 2 µm. (b) Four-probe resistance of the device. (c) Combined light reflection and photocurrent image (ICfA, scale bar ) 4 µm), measured with a 532-nm laser (230 µW) and a 60× objective with NA ) 0.7. (d) Schematic of the photocurrent generation in the graphene device.
induces localized generation of the PC, and its collection at different electrodes and gate potentials allows us to probe how the transport and collection of the PC is influenced by the electric potential profile of the graphene device. In addition, thermoelectric effects caused by laser-induced heating of electrodes can provide important information for identifying the majority carrier of SLGs in different conducting regimes. The SLG device used in our study is shown in Figure 1a. SLG is deposited by the exfoliation method on a 290-nmthick silicon oxide layer thermally grown on top of a silicon substrate, which served as a back gate.2 SLG, which is visible as the darker shade in Figure 1a, is then contacted by eight Cr/Au electrodes, as defined by electron beam lithography and liftoff. Conductance measurements confirm that all eight electrodes made an electrical contact to the SLG. Because there are eight electrodes, a variety of electrical configurations are possible. To specify the electrical current path, we designate each electrode with letters A through H, as shown in Figure 1a, and use the notation IRfβ, where R and β indicate the bias electrode and the electrode where the current was collected (ground), respectively. Finally, the distance between electrodes A and E is ∼12 µm, and that between B and H is 2.5 µm. Figure 1b shows the four-probe measurement of the electrical resistance of the graphene device, as a function of the gate bias VG. The resistance reaches the maximum of 2.2 kΩ at VG ≈ 12 V (VDirac). This corresponds to the electron band configuration in which the Dirac point of the graphene electron band is aligned with the Fermi level. As VG moves away from VDirac, the density of carriers in graphene increases, resulting in a decrease in resistance. From the conductance Nano Lett., Vol. 9, No. 5, 2009
plot, we measure an electron (hole) mobility of 0.42 (0.57) m2 V-1 s-1.19 In our device, VDirac drifts slightly over the course of days but remains constant during the course of each experimental run, as confirmed by a conductance measurement just before and after each experiment. A SPCM image of our device measured with a green laser (λ ) 532 nm) is shown in Figure 1c. The electrical and optical images, taken simultaneously, are shown overlaid on top of each other. The gray light-reflection image shows the positions of all eight electrodes, while the color scale displays the PC ICfA measured while only two electrodes (C and A) are connected to an external measurement circuit at zero bias. All the other electrodes were floating. In this image, a strong PC signal with alternating polarity is observed along the edges of several electrodes, most notably A, B, and C. The presence of the strong PC spots here is due to electron band bending and local electric fields located near electrodegraphene contacts, a behavior reported in detail in previous studies with carbon nanotubes and silicon nanowires.14,16 The mechanism for the PC signal along the dotted line in Figure 1c is depicted in Figure 1d. Upon laser illumination, electron-hole pairs will be separated and then transported to opposite directions due to local electric fields, resulting in a net electrical current. The polarity of this PC is dependent on the direction of the band bending. For instance, a positive PC will be measured with the laser beam near electrode A only if the graphene interior is more n-type than near A, as depicted in the upper image of Figure 1d. We note that a strong PC signal is observed near the edges of electrode B, with the polarity consistent with the PC signals near electrodes A and C, although B is not resistively connected to the external circuit. This is due to the fact that floating electrodes are still connected to external macroscopic conductors, including bonding wires and sample adapters, which are capacitively, although not resistively, coupled to the external ground. The electric potential near floating electrodes will be similar to the grounded electrodes. Therefore, a PC will be generated near floating electrodes in the same manner, regardless of the applied gate bias, although it can be collected only by resistively coupled electrodes. The intensity of the PC is strong and becomes as large as 0.7 µA under 0.23 mW of laser power. For this wavelength (λ ) 532 nm, 2.33 eV) and power, the maximum amount of PC will be 2.3 µA, assuming 2.3% absorption as measured by Nair et al.,10 and perfect charge separation. This indicates that more than 30% of the absorbed photons contribute to the PC near electrode contacts. Because the actual amount of light absorption by graphene is reduced due to the presence of the electrodes, we expect the real maximum efficiency for charge separation (and, thus, photocurrent generation) is significantly larger than observed here and should reach 40%-60%. Since the PC signal is correlated with local electron band bending, PC images taken at different gate biases can provide spatially resolved information regarding electron band alignment in SLG, even in a more-complicated multielectrode geometry. In Figures 2a and 2b, we show two PC images (IEfA), each taken at VG ) +10 V (n-type) and VG ) -10 1743
Figure 2. PC images (IEfA) measured at different gate biases using a 780-nm laser (0.34 mW) and a lock-in technique (10 kHz). Each measured at (a) VG ) 10 V (n-type) and (b) VG ) -10 V (p-type). (c) PC vs VG, measured along the dotted line in panel (b); inset shows the DC conductance measurement. (d) Schematic band diagram at different gate biases. (e) Plots of PC vs VG at positions 1 and 2.
V (p-type), while the six electrodes in the middle are floating. Therefore, electron-hole pairs in this case, once separated, can be collected only by electrodes E and A, although other electrodes can cause strong generation of the PC. Because we connected electrode E to the external ground, rather than electrode C here, it is possible to map out the PC near all the middle electrodes. Both images were taken with a laser at a wavelength of 780 nm with 10 kHz lock-in modulation. The most striking feature in these figures is the complete reversal of the PC polarity. This is consistent with the mechanism of PC generation discussed earlier. As the graphene changes from n-type (Figure 2a) to p-type (Figure 2b), the electron band bending direction changes from downward to upward as we move away from electrodes. The PC signals near floating electrodes in the middle also change direction in the same way. In Figure 2c, we plot the VG dependence of PC IEfA near several electrodes. Here, we scan the laser beam along the dotted line in Figure 2b as we slowly decrease VG from 10 V to -10 V. We clearly see that the PC near each electrode contact changes from n-type PC to p-type PC at a gate bias VFB, where each VFB value is denoted by the dots located at ∼2 V (electrode A-B), 3 V (electrode B-C), 4 V (electrode C-D), and 8 V (electrode D-E). These gate biases (VFB) of PC polarity reversal signify the flatband condition where the electron band near the contact is level (thus no electric field) as depicted in the middle of Figure 2d. While the PC reversal and the presence of the flatband points are both expected behaviors and were recently reported by Lee et al.,11 the variation of VFB from electrode to electrode is a surprising observation. VFB clearly increases 1744
monotonically as we move from electrode A toward electrode E. The same trend was observed when we perform a similar measurement along the line that crosses electrodes of the opposite side A-H-G-F-E. In an ideal graphene without sample impurities, it is expected that the Fermi level lies at the Dirac point. However, in the presence of surface impurities including oxide charges, adsorbates, and surface-induced graphene deformations,19 the local electric potential of graphene will vary and induce local variations of the electron (or hole) density, as recently reported to exist for a submicrometer length scale by Martin et al.20 In fact, such variation will exist at all length scales. The variation of VFB in our measurement strongly indicates that a similar potential variation does exist over the 10-µm length scale. Based on our measurements of ∆VFB for different electrodes discussed previously (∼2 V between electrodes A-B and C-D), we can estimate ∆nimp, which is the variation in the impurity charge density, to be ∼1.7 × 1011 cm-2 over the same region, assuming that ∆nimp (and the electrostatic potential associated with it) is the main cause for nonzero ∆VFB and that the gate capacitance per unit area is 1.2 × 10-4 F/m2. Although our measurement cannot elucidate the origin of this behavior in further detail, the observation of a large-scale potential variation is significant, because this potential will strongly affect the nonlocal electrical conductance measurements of graphene. Based on the data in Figure 2, we can also determine the electron band configuration at the contact. As shown in the inset to Figure 2c, the maximum resistance is observed at VG ) 7 V, a value that is larger than VFB for most electrode contacts (except electrodes D-E). This indicates that, for most electrodes, the electron band configuration is as shown in Figure 2d, where the local Dirac point in graphene is denoted by a gray line for different gate biases. Our data show that the Dirac point is located higher than the Fermi level at VFB, forming a p-type contact between graphene and electrodes. The formation of a p-type contact is shown more clearly in Figure 2e, where we measure the PC signal at positions marked by “1” and “2” in Figure 2a and scan VG. Both PC signals cross zero near VG ) 3 V and then change sign. However, the graphene is p-type up to VG ) 7 V, after which it changes to n-type. Based on the previous description, we expect the formation of a p-n junction near the majority of contacts when the gate bias exceeds 7 V. Moreover, the formation of local p-n junctions can explain the enhanced n-type PC in Figures 2c and 2e.12,14 The two-dimensional (2D) nature of graphene also provides a unique system where the spatial orientation of PC collection can be studied. In Figure 3, we describe an experiment where all eight electrodes were connected to an external ground, maintaining the same potential, while VG was tuned so that the graphene was kept n-type. Unlike the data shown in Figures 1 and 2, here, the PC generated by the laser illumination can be collected by any of the eight electrodes. Figure 3a shows eight PC images, each representing the PC collected by different electrodes. For instance, the top Nano Lett., Vol. 9, No. 5, 2009
Figure 3. PC collection in multielectrode geometry. (a) Maps of PC collection when all eight electrodes are connected for PC measurement (each image shows the PC collected at the corresponding electrode, as a function of the laser position); a 780-nm laser (0.34 mW) and a lock-in technique (10 kHz) were used (red denotes positive current). (b) The map of PC generation efficiency obtained by adding all absolute values from all eight PC images in panel (a). Scale bar ) 2 µm.
left image (IB) plots the PC collected at electrode B. Consistent with our previous measurements of PC when the graphene is n-type, it shows positive PC (shown in red) near electrode B, whereas all the other electrodes show a negative PC (blue). This negative PC, located adjacent to electrodes other than B, provides information about the distribution of PC generated in each position. More specifically, the blue spots in the IB image show the fraction of PC that is generated at each position and is collected by electrode B. By comparing all eight PC images shown in Figure 3a, we observe two distinct trends. First, most of the collected PC is generated on neighboring electrodes. For instance, the image of IB shows a strong PC signal only near the electrodes A, C, and H, which are the three electrodes that are closest to B. By comparison, the PC image of IH shows a strong signal near electrodes A, B, C, and G, which, again, are the four closest neighbors. Second, within the same electrode, the PC signal is stronger along the edge that faces directly toward the electrode where PC is collected. We can see this most vividly by examining the maxima of the negative PC signals around electrode C for the PC plots of IB, ID, and IG Nano Lett., Vol. 9, No. 5, 2009
(see circled areas in Figure 3). Electrode B collects more PC generated near the upper edge of electrode C, whereas electrode D collects more from the lower edge and electrode G colelcts more from the left edge of electrode C. This striking PC distribution can be qualitatively understood based on the mechanism for PC generation and the reported mean free path in SLG. In our graphene device, the PC is strong near each electrode contact, because of the electron band bending and the local electric fields. Because each electrode potential is constant over its edge, the electric field is perpendicular to the edge of the electrode and, hence, the photocarrier generated in graphene near an electrode contact will be pushed normal to the edge of the electrode, thereby gaining momentum along this direction. While this initial momentum will soon be lost after many scattering events in other materials, the average momentum will still be along the initial direction in SLG, because of a long carrier mean free path (as large as 0.4-0.6 µm),2,21 leading to maximum PC collection by the electrode positioned along this direction. In contrast, if we assume that PC collection is determined only by the relative resistance between a pair of electrodes (diffusive regime), the PC should be equally distributed among different electrodes, because the electrode contact resistances (typically on the order of 5 kΩ) are much larger than the graphene resistance (
