Influence of SiO2 or h-BN substrate on the room-temperature electronic transport in chemically derived single layer graphene

The substrate effect on the electronic transport of graphene with a density of defects of about 0.5% (0.5%G) is studied. Devices composed of monolayer 0.5%G, partially deposited on SiO2 and h-BN were used for transport measurements. We find that the 0.5%G on h-BN exhibits ambipolar transfer behaviours under ambient conditions, in comparison to unipolar p-type characters on SiO2 for the same flake. While intrinsic defects in graphene cause scattering, the use of h-BN as a substrate reduces p-doping.

Wet-chemically prepared graphene from graphite can be stabilized in solution by covalently bound oxo-groups using established oxidation protocols. [1][2][3] In general, the materials obtained are termed graphene oxide (GO). However, the chemical structure varies and the carbon lattice may even be amorphous due to the evolution of CO 2 during synthesis. 4 Thus, in this study we use oxo-functionalized graphene (oxo-G), a type of GO with a more dened structure, as proven in our previous work. 3 The oxygen-containing groups on the graphene basal plane and rims of akes and holes make GO a p-type semiconductor with a typical resistance of 10 10 -10 13 U sq À1 5,6 and a band gap of about 2.2 eV. 7,8 The reductive defunctionalization of GO leads to a certain type of graphene (G), oen named reduced GO (r-GO). 4,9 Removal of oxo-groups from the surface can be achieved by chemical reduction, 9,10 electrochemical methods, 11,12 electron beam treatment 13 and was observed in situ by transmission electron microscopy. 13 Thermal processing of GO instead leads to a disproportionation reaction forming carbon with additional vacancy defects and CO 2 . 14 In general, the reduction of GO turns r-GO from a semi-conductive material to a semi-metal. Mobility values were determined in eld effect transistor (FET) devices. 15,16 Generally, the quality of graphene strongly depends on the integrity of the hexagonal carbon lattice. Thus, mobility values of 10 À3 and up to 10 3 cm 2 V À1 s À1 were reported, 3,17,18 with the resistance uctuating between 10 3 and 10 6 U sq À1 . [19][20][21] We reported on the highest mobility values of chemically reduced oxo-G (with about 0.02% of lattice defects) of 1000 cm 2 V À1 s À1 , 3 determined by Hall-bar measurements at 1.6 K.
Hexagonal boron nitride (h-BN) has been proved to be an excellent substrate for matching graphene-based materials owing to its atomic atness, chemical inertness and electronic insulation due to a bandgap of $5.5 eV. 22 Up to now, most studies with graphene deposited on h-BN were restricted to measurements with virtually defect-free graphene. 23 To the best of the authors knowledge, no studies reported transport measurements based on single layers of GO or oxo-G on h-BN substrates. No studies are reported with graphene derived from GO or oxo-G on single-layer level. Recently, we found that chemical reactions can be selectively conducted close to the rims of defects. 24 However, before functionalized devices can be studied, the lack of knowledge on the ambient environment device performances of graphene with defects and the inuence of substrates must be addressed. Therefore, we fabricated the devices composed of 0.5% G, partially deposited on SiO 2 (SiO 2 / 0.5% G) and h-BN (h-BN/ 0.5% G) (Fig. 1). Areas of the same ake on both materials are used to ensure reliable measurements and to prove that the results stem from the inuence of the substrate rather than from the difference between devices. Thereby, the 0.5% G exhibits an I D /I G ratio of about 3-4, corresponding to 0.5% of defects, according to the model introduced by Lucchese and Cançado. [25][26][27][28] Our results demonstrate that the h-BN layer is responsible for a downshi of the Dirac point and a more narrow hysteresis, resulting in ambipolar transfer behaviours in h-BN/ 0.5% G.

Results and discussion
To gain structural information of 0.5% G, akes of 0.5% G were deposited on HOPG surface and examined by scanning tunnelling microscopy (STM) under an ultra-high vacuum (10 À10 mbar). The average height of a single 0.5% G ake was determined as 0.9 nm (Fig. 2a). At higher resolution, two different morphologies are detected in the 0.5% G ake, as depicted in Fig. 2b. The atomically resolved structure is assigned for the dark region while the resolution fades away in the bright region. The diffraction spots marked in dashed red indicate the hexagonal lattice of graphene in the dark regions, shown in Fig. 2c. The bright regions are caused by the oxygencontaining groups attached to the carbon lattice, which breaks the lattice periodicity of graphene and subsequently lead to no apparent diffraction feature in the reciprocal space (Fig. 2d).
Atomic scale electronic properties of 0.5% G were explored using scanning tunnelling spectroscopy (STS). Fig. 2e displays the I (V) spectrum of the 0.5% G surface. Compared to the tunnelling current at the dark region, there exists an apparent suppression of current at the bright region owing to a lower conductivity in the distorted graphene lattices. For the averaged I (V) spectra of the whole area, the metallic behaviour of the  spectrum (averaged over >100 single spectra) and corresponding dI/dV curves recorded at the dark area (red curve), bright area (blue curve) and whole area (black curve), respectively. 0.5% G ake is found. This phenomenon is also conrmed by the differential conductivity (dI/dV) curves in Fig. 2f. The Dirac point is determined from the minimum value in dI/dV curves. The Dirac point in dark region is located at 0.0 V, suggesting low impurity-related doping level. In contrast, the bright regions exhibit a positive shi of the Dirac point of about 50 mV, likely due to the presence of oxygen groups. For the entire scanned areas, the 0.5% G ake exhibits a p-type electronic doping feature with the average Dirac point at about 20 mV.
For the fabrication of the heterostructure of h-BN/ 0.5% G or SiO 2 / 0.5% G, akes of oxo-G were rst deposited on SiO 2 substrate by Langmuir-Blodgett technique, 29 as shown in Fig. 3a. Then 0.5% G akes were prepared by reduction using vapor of HI/TFA (in inset of Fig. 3b). 30 The h-BN akes used in this study were exfoliated from h-BN single crystals. 31 Next, the heterostructures of h-BN/ 0.5% G or SiO 2 / 0.5% G were prepared by a dry transfer technique. 32 Fig. 4a shows an AFM image of a h-BN/ 0.5% G heterostructure, which consists of SiO 2 substrate with multilayer h-BN ake and a monolayer 0.5% G ake ($25 Â 10 mm 2 ) partially covering the h-BN. The AFM image in Fig. 4b, obtained within the marked area in Fig. 4a, revealed that the transfer process induced wrinkles and folds in 0.5% G. The height prole of the single 0.5% G ake on SiO 2 is shown in Fig. 4c (compare Fig. S1 †) and depicts a thickness of about 2 nm. This height is much thicker than 0.9 nm measured by STM for similar monolayer 0.5% G on HOPG.
A major plausible reason is that e.g. water molecules are inevitably adsorbed on the hydrophilic SiO 2 surface (treated by O 2 plasma) leading to an approximately nanometer-thick hydrogen-bonded water layer and cleaved oxo-groups captured between SiO 2 and 0.5% G. 33 In contrast, although small amounts of polymer residues are likely trapped between h-BN and 0.5% G, the measured thickness of the same 0.5% G ake on h-BN is   $1 nm as shown in Fig. 4c, which is almost the same result as the thickness determined by STM. The 0.5% G ake on $6 nm thick h-BN (Fig. 4c) possesses a lower roughness ($0.5 nm) than on SiO 2 ($1 nm). Therefore, h-BN, as a passivation layer, can not only negate the inuence of trapped water on graphene, but also improves accuracy in the AFM thickness measurements of monolayer 2D akes.
Average Raman spectra of the 0.5% G supported by SiO 2 and h-BN, respectively, are shown in Fig. 5a. The primary peaks are the D peak near 1340 cm À1 , the G peak near 1555-1557 cm À1 and the 2D peak near 2667 cm À1 . The D peak of 0.5% G on each interface is mainly activated by defects in the carbon skeletons. The G and 2D peaks closely relate to the quality of graphene. The almost unchanged positions of the three peaks indicate that wrinkles and residual polymers induced during the transfer processes do not produce obvious doping effect on the single layer 0.5% G. We use scatter plots of I D /I G versus G 2D to further conrm the quality of the 0.5% G in Fig. 5b. For the 0.5% G on h-BN, the I D /I G ratio is about 3.3, within the standard deviation of the I D /I G ratio of 3.1 determined on SiO 2 . Based on the model introduced by Lucchese and Cançado et al., 25,26 the density of lattice defects is related to 0.5% for the devices on h-BN and SiO 2 . This density of defects relates to the average distance between defects of around 3 nm. The related defect density (n D ) is 4.0 Â 10 12 cm À2 on h-BN and SiO 2 , respectively, calculated from the equation n D (cm À2 ) ¼ 10 14 /(pL 2 D ). 25 The G 2D of the Raman 2D band is sensitive to the presence of defects. For the monolayer 0.5% G on h-BN, only a slightly smaller G 2D of $70 cm À1 is observed than on SiO 2 ($72 cm À1 ). The same monolayer 0.5% G, partially deposited on SiO 2 and h-BN, presents almost the same G 2D . Therefore, the quality of the investigated ake is the same on SiO 2 and h-BN, respectively.
Reference experiments to determine the contact resistance were conducted using four-probe measurements. The surface resistance is determined to roughly 21 kOhm in four-probe conguration and 23.5 kOhm in two-probe conguration (Fig. S2 †). Thus, further investigations were conducted in twoprobe conguration under ambient conditions. For our transport measurements, we prepared one device with monolayer 0.5% G on SiO 2 substrate (Fig. S3 †), two devices with the same monolayer 0.5% G ake that are in part on SiO 2 and on h-BN ( Fig. 1 and S4 †) and one device with monolayer 0.5% G on h-BN substrate (Fig. S5 †). The patterning of the electrodes was achieved by standard electron beam lithography processing and subsequent deposition of 5 nm Cr/70 nm Au by thermal evaporation. The electrical performance of the 0.5% G ake on h-BN and SiO 2 , respectively, is summarised in Table 1. The resistance of 0.5% G on h-BN and SiO 2 measured at V bg ¼ 0 V ranges a Reference device of 0.5% G on SiO 2 , see Fig. S3 (channel: 1-2). b Reference device of 0.5% G on SiO 2 , see Fig. S3 (channel: 2-3). c Reference device of 0.5% G on SiO 2 , see Fig. S4 (channel: 1-2). d Reference device of 0.5% G on h-BN, see Fig. S4 (channel: 3-4). e Reference device of 0.5% G on h-BN, see Fig. S5 (channel: 1-2). widely, from 5.0 kU to 34.4 kU. But the resistances are signicant lower compared to >10 6 U reported for similar devices. 17 Transfer curves (I ds -V ds ) of 0.5% G on h-BN is shown in Fig. 6a. The Dirac points are located at around +20 V. The hysteresis effect of the 0.5% G on h-BN is observed in ambient environment for sweeping continuously from À50 to 50 V in forward direction and then back to À50 V (backward direction). From the red dashed lines presented in Fig. 6a, a room-temperature hole mobility (m h ) of 5.6 cm 2 V À1 s À1 is extracted using the equation m ¼ (L/W) Â (1/(C ox V ds )) Â (dI ds /dV bg ), 34 where C ox ¼ 1.15 Â 10 À8 F cm À2 . As the output curves (I ds -V ds ) exhibit ohmic behaviour (Fig. 6b) we conclude that there is no Schottky contact between 0.5% G and metal electrodes. For the 0.5% G deposited on the overlapped SiO 2 -h-BN hetero-substrate (transport measurements performed between electrodes 2 and 4, shown in Fig. 1c), we observe only p-type character of the I ds -V ds curves with the Dirac point shied to about +30 V (Fig. 6c).
In contrast to 0.5% G on h-BN and overlapped SiO 2 -h-BN hetero-structure, the 0.5% G on SiO 2 exhibits unipolar p-type character (Fig. 6e). The point of the minimum conductivity in the I ds -V bg curve is not observed and the Dirac point moves to higher positive voltage (>43 V). Obviously, electrical transport of the 0.5% G on SiO 2 is completely governed by holes with hole mobility m h estimated to about 11.6 cm 2 V À1 s À1 . In addition, the I ds -V bg curves exhibit an increase of hysteresis in SiO 2 / 0.5% G device with a shi of V bg (DV bg z 7.3 V) between the forward and reverse sweeps, compared to the h-BN/ 0.5% G device with DV bg z 2.6 V. Substrate change from h-BN to SiO 2 induces trapped holes with density higher than 1.6 Â 10 12 cm À2 using Dn t ¼ DV Dirac point (C ox /q), 2 where q is the elementary charge, DV Dirac point > 43-20 ¼ 23 V. In general, a high density of charge traps can cause hysteresis and lead to reduced mobility of graphene samples. 35 However, as summarized in Table 1, mobility values on SiO 2 are higher and the resistance is lower than on h-BN. The main reason for that contradictory nding is that for 0.5% G defects are the dominant scatterers reducing the carrier mobility. This is consistent with Raman results of Fig. 5b. As further reference experiments we conducted transport measurements of defective graphene, here 0.8% G on SiO 2 . As shown in Fig. S6, † due to the higher density of defects the hole mobility values are 0.6 cm 2 V À1 s À1 in ambient and 0.9 cm 2 V À1 s À1 in vacuum. However, the Dirac point shis only from 60 V in ambient to 30 V in vacuum. Those results are in agreement with the STS measurements, which indicate p-doping of 0.5% G in vacuum. It could however be expected that oxo-groups with ÀI and ÀM effects, 2,3 decorating the rims of vacancy defects, may be responsible for trapping hole carriers. However, the experimental results, such as transport and AFM measurements, give evidence that p-doping is strongly induced by the SiO 2 substrate and cleaved oxo-species, such as water or organosulfate, which are trapped between SiO 2 and 0.5% G. Therefore, based on the AFM height determination on SiO 2 , the knowledge about the chemical structure and the reduction mechanism of oxo-G to 0.5% G we propose that molecules, such as water or hydrogensulfate stemming from oxo-G (Fig. 7a) are trapped between the SiO 2 substrate surface and 0.5% G (Fig. 7b). In comparison, h-BN is affected by the local polarity of h-BN/ 0.5% G. As a result, spurious dopant molecules may get squeezed out (Fig. 7c), as is also supported by the measured height and roughness results determined by AFM.
Conclusions 0.5% G is a p-doped material and defects determine the scattering of charge carriers. Using h-BN as substrate leads to less trapped molecules, which are responsible for p-doping. In this regard, most likely hydrogen-bonded water and other cleaved oxospecies are captured between SiO 2 and 0.5% G causing pdoping, as a consequence of chemical reduction of oxo-G. The ambipolar behaviour with V Dirac point of +20 V was therefore observed for the h-BN/ 0.5% G structure while unipolar p-type response was shown for the same 0.5% G ake on SiO 2 . Transfer characteristics show a reduction of hysteresis in the h-BN/ 0.5% G. The mobility of the SiO 2 / 0.5% G is determined to 7.4- Fig. 7 Proposed model of trapped species upon cleavage of oxogroups upon reduction and influence of substrate. (a) Chemical sketch of the structure of oxo-G with the graphene lattice decorated by hydroxyl-, epoxy-and organosulfate groups. (b) 0.5% G prepared by chemical reduction of oxo-G; covalently bound oxo-groups are cleaved and at least partially trapped between 0.5% G and the SiO 2 substrate. (c) 0.5% G on h-BN; cleaved oxo-groups may not be trapped between h-BN and 0.5% G because they are squeezed out.

Conflicts of interest
There are no conicts to declare.