Towards electron transport measurements in chemically modified graphene: The effect of a solvent

Chemical functionalization of graphene modifies the local electron density of the carbon atoms and hence electron transport. Measuring these changes allows for a closer understanding of the chemical interaction and the influence of functionalization on the graphene lattice. However, not only chemistry, in this case diazonium chemistry, has an effect on the electron transport. Latter is also influenced by defects and dopants resulting from different processing steps. Here, we show that solvents used in the chemical reaction process change the transport properties. In more detail, the investigated combination of isopropanol and heating treatment reduces the doping concentration and significantly increases the mobility of graphene. Furthermore, the isopropanol treatment alone increases the concentration of dopants and introduces an asymmetry between electron and hole transport which might be difficult to distinguish from the effect of functionalization. The results shown in this work demand a closer look on the influence of solvents used for chemical modification in order to understand their influence.


Introduction
Graphene is an electronic material with high electron mobilities even at room temperature [1]. Usually graphene is prepared by exfoliating individual layers from bulk graphite and putting them down on a substrate [2]. With such techniques it has become possible to prepare samples displaying quantum Hall effect, testifying to the high electronic quality of such systems [3,4]. Further improved mobilities were achieved by suspending graphene flakes [5], or, very recently, by depositing graphene on boron nitride [6]. It is generally believed that unintentional adatoms on top of the graphene flake and charge traps in the substrate limit the mobility for conventional devices [7,8,9].
Chemical modification of graphene has been achieved by a number of methods and has been investigated by Raman measurements and transport studies [10,11,12,13]. Applying chemistry on graphene changes the local carbon-carbon bond structure, the orbitals and hence the electronic properties of the material. Until now, it is not so clear how conventional methods used in almost any graphene sample preparation, such as baking in inert gas atmosphere in combination with rinsing in water or organic solvents affect the electronic quality of a graphene system. Such treatments are also standard conditions in chemical reactions and can induce a change on electron transport along with chemical functionalization itself. Therefore solvent effects should be taken into account when analyzing transport data of chemically derivatized graphene samples.
In the first part of this work we present a confocal Raman spectroscopy analysis of graphene chemically modified with aromatic diazonium ions. A difference in reactivity between single layer, bi-layer and single layer edge is observed. In the second part we first show the influence of functionalization on the electronic transport properties of graphene, and afterwards we focus on the influence of repeated treatment with baking and rinsing in isopropanol. Here we find that the treatment with only isopropanol leads to an increase in the doping concentration and an asymmetry between electron and hole transport which is partly similar to the effect of the functionalization. In addition we observe that the combined treatment with isopropanol and baking leads to a higher electronic quality than just heating alone. This is further investigated at low temperatures in the last part of this paper.

Experimental method
Single and bi-layer graphene flakes were exfoliated from natural graphite and deposited onto a Silicon substrate covered by ≈285 nm thermal silicon dioxide [2] and identified using Raman spectroscopy and light microscopy [14,15].
For the Raman spectroscopy study the chemical functionalization is carried out at room temperature by immersing the chip into a 20 mmolL −1 solution of water-soluble nitrobenzene diazonium salt (4-nitrobenzene diazonium tetrafluorborate from Sigma Aldrich) [12]. After the functionalization the chips were cleaned once in isopropanol (1 min), two times in water (1 min), a second time with isopropanol (1 min) and finally blown dry with nitrogen. For the electronic transport experiments the chemical functionalization was carried out at 0 • C using a 4 mmolL −1 solution. The cleaning procedure after the functionalization was the same as for the Raman spectroscopy study. It should be pointed out that due to the different reaction conditions described above the Raman spectroscopy study and the transport study cannot be directly compared in terms of amount of induced disorder as a function of reaction time.
During the Raman spectroscopy study laser power was kept at 2 mW in order to avoid heating and the introduction of defects due to the laser.
For the electronic transport experiments Ohmic contacts were defined on the graphene flakes using standard electron beam lithography techniques followed by the evaporation of Cr/Au (2/40 nm). The highly doped silicon substrate is used as a global gate to tune the overall Fermi energy of the device. The Hall bar used to investigate the influence of isopropanol and heating treatment on the transport properties of graphene is patterned in a second electron beam lithography step followed by reactive ion etching.

Raman spectroscopy of chemically functionalized graphene
Raman spectroscopy is a powerful tool for both identifying the number of graphene layers [14,15] as well as for monitoring doping [16,17], defects [18] and chemical functionalization [12,13] of graphene. The most prominent features in the Raman spectra of graphene are the G band (around 1580 cm −1 ) and the 2D band (around 2700 cm −1 ). In addition, in the presence of defects or at the edge of graphene the disorder induced D-line located around 1350 cm −1 can be observed [19]. Here we functionalize graphene using diazonium chemistry and monitor the introduction of defects in graphene lattice by measuring the intensity of the D-line of the Raman spectra.
Diazonium chemistry has previously been used to functionalize a variety of carbon forms [20,21,22,23,24] and it was recently shown that also graphene can be functionalized in a similar manner as the other carbons forms using the same chemistry [12,13]. In this experiments a flake is used that has both single and bilayer domains, allowing the direct identification of differences in the chemical reactivity towards the diazonium reagent. In Fig. 1(a) the 2D map of the integrated D-peak intensity is shown after 20 minutes immersion in the reaction medium. Bright areas correspond to high intensity and dark areas to low intensity. Three distinct domains with similar intensities are visible, which can be attributed to bi-layer, single layer or single layer edge of the graphene flake, respectively. This is further shown in Fig. 1(b) where the intensity of the D-peak along the white line in Fig.1(a) is plotted.
The D-peak intensity of the Bi-layer region is very hard to identify, as the signal overlaps with adsorbed species [12]. It has been shown by Strano et al. [13] that long reaction time and extensive washing procedure is necessary to identify a small D-peak on Bi-layer graphene after functionalization with diazonium ions. On single-layer graphene  the D-line integral is significantly higher than on bi-layer. The higher reactivity towards diazonium chemistry for single-layer than bi-layer has been attributed to less ripples on the bi-layer surface [12] and to screening of electron-hole puddles in bi-layer [13]. Furthermore the single layer part can be divided in two regions of distinct intensities, bulk single layer with a lower D-peak intensity and edge single layer with a higher intensity. This increased edge intensity was also shown earlier and is attributed to a higher reactivity of the reagents towards the edge, due to a higher degree of flexibility, which is necessary to change the local geometry from planar sp 2 to tetrahedral sp 3 . In addition it was recently shown that these edge regions grow over the whole single layer area with prolonged reaction time [12]. This is an indication that near defects or functional groups on the surface, the carbon atoms react more easily with the diazonium reagents. In Fig.1(c) the difference in reactivity between the different parts of the flake is schematically illustrated. For a detailed investigation of the dependence of disorder as function of exposure time to the reaction medium see Koehler et.al [12].

Room temperature transport measurements
The possibility of controlled doping and, as investigated above, selective functionalization of graphene edges makes chemical modification of graphene interesting for electronic transport experiments. In Fig. 2(a) room temperature measurements of the conductance (G) as a function of backgate voltage (V BG ) are shown for (i) an unfunctionalized sample (only heated in order to remove dopants from the surface), (ii) after 5 minutes of functionalization and (iii) after 100 minutes of functionalization. These are two-terminal measurements on an unpatterned graphene flake (see the light microscope image of the measured device in the inset in Fig. 2). From Fig. 2(a) it can clearly be seen how the functionalization leads to an increased p-doping of the graphene flake. Before functionalization the point of minimum conductance (the Dirac point, V DP ) is located at +9 V in backgate. After 5 minutes of functionalization the Dirac point is shifted to +21 V and after 100 minutes of functionalization the Dirac point is at +31 V. In Fig. 2(b), where the backgate traces are normalized with respect to V DP , it can be seen that the functionalization introduces a small asymmetry between electron and hole transport. This asymmetry is much weaker than observed previously by Farmer et al [10]. In addition it can be seen from Fig. 2(b) that the mobility (slope of G versus voltage) of the graphene flake is not significantly changed after functionalization. Both observations can be explained by a lower amount of functionalization due the low temperature (0 • C) used in this work.
In the functionalization process described above the graphene flake is first immersed in water containing the reactive diazonium ions and afterwards in isopropanol to remove unreacted species and improve the drying step. In order to analyze the influence only of the chemical functionalization on the electronic transport properties of graphene it is crucial to know the effect of the involved solvents. In the following we will therefore investigate the influence of isopropanol and baking on graphene's transport properties,  which are part of the chemical and physical treatments involved in the reaction and measurement process.
To investigate the influence of isopropanol and heating we use the Hall bar shown in the inset in Fig. 3(a). The width of the Hall bar is ≈1 µm and the length between two voltage probes is ≈2 µm. All following measurements are four-terminal measurements. For the isopropanol treatment the chip with the Hall bar is immersed in isopropanol for 5 minutes and afterwards blown dry with nitrogen gas. The heating of the sample is done in the sample holder while the vacuum is constantly pumped. In order to monitor changes in the conductivity of the sample during the heating a constant current of 10 nA is applied to the Hall bar and the four-terminal resistance is measured at V BG = 0 V. The sample is always heated at 150 • C until the measured resistance is stable. This may take many hours.
In Fig. 3(a) the conductivity (σ) of the Hall bar as a function of applied V BG for (1) the untreated sample, (2) after heating the sample, (3) after treating the sample with isopropanol and (4) after heating the sample again is plotted. It can be seen that both the mobility and the position of the Dirac point is changed significantly after the different treatments.
For the untreated sample the Dirac point is located at +43 V. The extensive doping of the pristine sample is probably due to resist residues and other dopants accumulated during the processing steps. In order to remove these dopants we always bake our samples before starting measurements (as we also did before functionalization).
Here it can be seen that after the initial baking of the sample the Dirac point has moved to +26 V. The corresponding change in mobility will be discussed below. As a next step we treat the sample with isopropanol. Fig. 3 shows that the Dirac point is shifted from +29 V to +34 V in backgate after the isopropanol treatment, which  means that isopropanol significantly p-dopes graphene. From Fig.3(b), where the traces before and after isopropanol treatment from Fig. 3(a) are normalized with respect to V DP , it can in addition be seen that the isopropanol introduces a strong asymmetry between the electron and hole conductivities. Above it has been shown that in the absence of significant sp 3 hybridization of the graphene surface, functionalization with diazonium salt does not lead to a suppression of conductance, only a shift of the Dirac point to more positive backgate voltages. The observed asymmetry after isopropanol treatment is larger than observed after the functionalization (Fig. 2(b)). However, it is similar to the asymmetry found by Farmer et. al after functionalization [10]. The qualitative similarities between the changes in the conductivity of graphene after isopropanol treatment and the changes observed after functionalization suggest that with the functionalization procedure described above it might be difficult to separate the effects of the diazonium salt and the effects of isopropanol.
In the final step we heat the sample a second time in order to see if we can remove the dopants introduced by the isopropanol treatment. Surprisingly the Dirac point does not only shift back to +29 V where it was located before the isopropanol treatment, it shifts much further to +8 V. Together with the corresponding increase in mobility this suggests that the electronic quality of graphene can be improved by repeated isopropanol treatments followed by heating. In case of the measurements (1) and (3) it is difficult to extract the electron mobilities and thus only the hole mobilities will be compared in the following. For the untreated graphene flake (1) we obtain a hole mobility of 2100 cm 2 /Vs. After the first heating step (2) the mobility has increased to 2700 cm 2 /Vs. The following isopropanol treatment (3) increases the mobility further to 3600 cm 2 /Vs and after the last heating (4) the mobility reaches 4700 cm 2 /Vs. Generally we expect the introduction/removal of dopants to decrease/increase the mobility. Here, after the isopropanol treatment, an increase in hole mobility is observed together with an increased doping. This might be due to the removal of some dopants and the introduction of a different kind of dopants.
The fact that annealing the sample removes dopants and improves the mobility is generally accepted. Therefore baking is normally a part of standard processing procedures for graphene. However, that a subsequent treatment with isopropanol followed by annealing is removing even more dopants has to our knowledge not been noted so far. We observe here that the repeated treatment with isopropanol followed by heating improves the quality of the sample far beyond the improvement due to the first heating.
In addition to the observed increase in sample quality after the combination of isopropanol treatment and heating the effect of the isopropanol treatment alone should also be pointed out. Isopropanol treatment alone leads to an increased pdoping and electron-hole asymmetry. These two effects are partly seen after chemical functionalization as well and thus it is important to be very careful when assigning shifts of the Dirac point and changes in electron-hole symmetry solely to the introduction of the modifying species.
In order to evaluate the connection between the functional groups and the transport experiments an estimate of the mean distance between the defects induced by the functional groups is necessary. This may be possible by evaluating the Raman data as shown by Lucchese et.al [25]. However our Raman data and transport data are from two different measurement cycles on different samples. Simulations showing the connection of defect spacing and transport in graphene nanoribbons have been shown by Lopez-Bezanilla et.al [26]. For further investigation of functionalized graphene and the influence of different solvents it would be therefore be favourable to perform Raman spectroscopy studies parallel with transport studies in order to make a more quantitative study about the defect density.

Low temperature transport measurements
The quantum Hall effect and the corresponding magnetooscillations of the longitudional resistance is found in two-dimensional systems of high quality and at low temperatures. The quality of the quantum Hall effect is a direct measure for the quality of the electronic system. Therefore, in order to further investigate the influence of the isopropanol treatment and confirm the improvement of the electronic quality of the graphene, we perform transport measurements in magnetic field at T =4 K. Fig. 4 shows the four-point longitudional resistance (R xx ) of the flake as a function of V BG at fixed magnetic field B =5 T after the first time heated (2) and after isopropanol treatment and the second heating (4). ((2) and (4) corresponds to Fig. 3). Before the isopropanol treatment R xx does not go to zero and only a weak splitting of the main resistance peak is observed. In contrast, after the isopropanol treatment and heating  R xx is clearly zero for filling factor ν =2 and in addition several more oscillations in R xx are visible. These measurements show that the electronic quality of the graphene flake is indeed improved after treating it with isopropanol and heating it.

Conclusions
To conclude we have presented confocal Raman spectroscopy studies of chemically functionalized single and bi-layer graphene and shown that the reactivity of the edges and the single layer parts are larger than the reactivity of the bi-layer parts. Furthermore we have performed a transport study of chemically modified graphene and found that the influence of an isopropanol treatment is comparable to the influence of the functionalization itself. It is shown that on one hand isopropanol leads to a p-doping similar to the p-doping observed after functionalization. In addition it is observed that isopropanol treatment followed by heating significantly improves the electronic quality of graphene beyond the improvement due to heating alone.

Acknowledgements
The authors thank Prof. Christofer Hierold for access to the confocal Raman microscope, T. Ihn for helpful discussions and the Swiss National Science Foundation for financial support.