Spectroscopic Studies of the Physical Origin of Environmental Aging Effects on Doped Graphene

The environmental aging effect of doped graphene is investigated as a function of the organic doping species, humidity, and the number of graphene layers adjacent to the dopant by studies of the Raman spectroscopy, x-ray and ultraviolet photoelectron spectroscopy, scanning electron microscopy, infrared spectroscopy, and electrical transport measurements. It is found that higher humidity and structural defects induce faster degradation in doped graphene. Detailed analysis of the spectroscopic data suggest that the physical origin of the aging effect is associated with the continuing reaction of H2O molecules with the hygroscopic organic dopants, which leads to formation of excess chemical bonds, reduction in the doped graphene carrier density, and proliferation of damages from the graphene grain boundaries. These environmental aging effects are further shown to be significantly mitigated by added graphene layers.


I. INTRODUCTION
Among many technological promises of graphene, 1-3 the feasibility of employing graphene as transparent conducting electrodes in optoelectronic devices has been demonstrated. [4][5][6] In particular, the application of graphene-based electrodes to organic photovoltaic cells (OPVCs) and organic light emitting diodes (OLEDs) are especially appealing because of the potential of achieving inexpensive and flexible optoelectronic devices. 6 In the case of graphene-based OPVCs, reasonable power conversion efficiency has been realized. 7,8 However, apparent degrading performance with time has been a major challenge for practical applications. 6 It has been reported that stacked multi-layers of graphene with a spacing less than 0.7 nm atop organic molecules could protect the OPVC from rapid decrease in the power conversion efficiency within the first three days, 6 and that coating of multi-layers of graphene on Cu or Ni could prevent the metal from rapid oxidation. 9 Moreover, stacked graphene oxide layers were found to be water impermeable. 10 Nonetheless, the underlying causes and reliable remedies for the aging effects of doped graphene have not been systematically investigated.
Here we report systematic imaging and spectroscopic studies of both hole and electron doped graphene as a function of time for up to 30 days under different conditions, and compare these results with those of undoped, pristine graphene. We find that the aging effect is more severe if the ambient humidity is higher, and that faster aging effects occur along the defects in graphene. Moreover, organic molecules intercalated between two sheets of monolayer graphene appear to be much more stable than those covered under a monolayer graphene sheet. Our experimental findings can be consistently explained by attributing the environmental aging effect to continuing reaction of H2O molecules with the hygroscopic organic dopants, which lead to propagating damages along the graphene grain boundaries and continuous decrease in the graphene carrier densities due to the degradation of organic dopants.

II. EXPERIMENTAL
We conducted time-dependent spectroscopic and imaging studies of doped graphene samples prepared under six different conditions, as summarized in Fig. 1(a). In addition, controlled pristine graphene samples without organic doping were studied as a function of time for comparison. The monolayer graphene films used for our investigation were (8×8) mm 2 in size and were synthesized by the chemical vapor deposition (CVD) process on copper foils, all purchased from Graphene Supermarket. The graphene films were transferred from their original copper substrates by a polymer-free technique, 11 as schematically shown in Fig. 1(b). Schematics of preparation procedures of the doped graphene samples, where doped monolayer graphene is prepared by following the "Target 1" process, whereas intercalated bilayer graphene is prepared by following the "Target 2" process. (c) The molecular structures for the hole-dopant PEDOT and the electron-dopant PEIE.
To prepare doped monolayer graphene, a graphene-on-copper sample was first placed in the middle of a graphite confinement and immersed in the ammonium persulfate etchant (Alfa Aesar, 0.2M) to etch away the copper substrate. After the copper foil was removed, the pristine graphene film was buoyed up by an aqueous solution (a mixture of deionized water and IPA in the ratio of 10:1) that replaced the etching waste. Meanwhile, a dilute PEDOT:PSS solution (Heraeus, CleviosTM PH 1000) for holedoping, or a dilute PEIE solution (Alderich, 0.05wt%) for electron-doping, was gradually introduced into the aqueous solution to obtain a doped graphene by liquid phase diffusion beneath the graphene sheet.
(See Fig. 1(c) for the molecular structures of PEDOT:PSS and PEIE.) Then the doped solution was completely rinsed off by the IPA-deionized water mixture prior to placing a substrate under the doped graphene to ensure that the substrate was free of residue. Finally, the IPA-deionized water mixture was pumped out to land the doped graphene onto the "target" (typically a SiO2/Si substrate), and the doped graphene was subsequently heated at 60 C for 10 minutes to improve the adhesion and to eliminate residue moisture.
To obtain the intercalated bilayer with the dopant incorporated between two graphene sheets, the aforementioned polymer-free transfer and doping process was first carried out with a monolayer grapheneon-copper as the target, which led to the graphene-dopant-graphene-copper assembly. The copper was then removed and another substrate (e.g. SiO2/Si) was introduced so that the final sample became intercalated bilayer graphene on SiO2/Si, as shown in Fig. 1 Raman spectroscopy, x-ray photoelectron spectroscopy (XPS) and ultraviolet photoelectron spectroscopy (UPS) were performed on all graphene samples. The Raman spectra were taken with a Renishaw M1000 micro-Raman spectrometer system using a 514.3 nm laser (2.41 eV) as the excitation source. A 50× objective lens with a numerical aperture of 0.75 and a 2400 lines/mm grating were chosen during the measurement to achieve better signal-to-noise ratio. For each sample after a given aging time, the Raman spectra were taken at nine different locations. The resulting spectra were analyzed, and all spectral characteristics were obtained by using the average values as the mean and the range of variations as the empirical error. XPS was performed under 10 9 Torr with a Surface Science Instruments M-Probe that utilized Al K  X-rays and a hemispherical energy analyzer. The UPS experiments were carried out in a Phi5400 system with a base pressure of 10 10 Torr. During the UPS measurement with He II (photon energy h = 40.8 eV) as the excitation source, the photoelectrons were collected by a hemispherical analyzer with an energy resolution of 0.05 eV. Time-evolved scanning electron microscopic (SEM) images were also taken on all graphene samples by a FEI Nova 600 SEM system with the following parameters: acceleration voltage = 5 kV, beam current = 98 pA, and working distance ~ 5 mm.
Additionally, Fourier transformed infrared (FT-IR) spectroscopic studies were carried out on PEIE using an FT-IR spectrometer (Nexus 470). Both transmittance and absorbance spectra were obtained over the spectral range from 400 cm 1 to 4000 cm 1 .
For studies of the aging effect in air, we verified that the averaged humidity in Pasadena during the period of our experiments varied from 21% to 34% with a mean value of 28% according to the official data from the National Oceanic and Atmospheric Administration. All "in-air" graphene samples were stored in an air-condition laboratory where the humidity was lower and varying less than the external humidity, ranging from 18% to 28% with a mean value of 23% over the same period of time. These samples were only exposed to external air briefly whenever they were transported to another building on campus for XPS and UPS measurements. Therefore, the humidity values for the "in-air" condition may estimated at ~ (23±5)%. To investigate doped graphene with saturated humidity, a CSZ MicroClimate environmental testing chamber was used for sample storage to simulate a saturated humidity condition, and the accuracy of controlled humidity was found to be ~ (95±5)% for ambient temperatures maintained at (25±3)C.

III. RESULTS & ANALYSIS
The electrical transport properties of the as-doped graphene and pristine graphene samples were characterized using the field-effect-transistor (FET) configuration at room temperature, confirming excess holes (~ 3.24 ×10 12 cm 2 ) in the PEDOT:PSS-doped graphene and excess electrons (~ 2.31 ×10 12 cm 2 ) in the PEIE-doped graphene, as shown by the gate voltage shifts in Fig. 2 Table I. Additionally, the spectral characteristics of as-grown graphene are included in Table I for In general, all doped graphene samples exhibit the following common time evolution regardless of the conditions. First, the 2D-band that represents a double-resonance process of a perfect monolayer graphene exhibited steadily decreasing peak intensity and increasing linewidth with time. Second, the peak position of the G-band, which is associated with the doubly degenerate zone-center E2g mode of graphene, downshifted with time while the intensity remained the same. Third, the intensity of the Dband, which is associated with the zone-boundary phonons due to defects, increased strongly with time.
Fourth, the full-width-half-maximum (FWHM) linewidth of all three Raman modes increased steadily with time.
To better understand how varying conditions influenced the aging of graphene, we normalize the following time-dependent spectral characteristics to their initial "as-doped" values: the peak positions of the 2D-and G-bands, the FWHM linewidths of the 2D-and G-bands, and the intensity ratios (I2D/IG) and (ID/IG) of 2D-to-G and D-to-G, respectively. Our rationale for choosing these quantities is that increasing FWHM linewidths of the 2D-and G-bands, decreasing values of (I2D/IG), and increasing values of (ID/IG) are generally indicative of the degradation of monolayer graphene quality. 16,17 Specifically, the magnitude of (ID/IG) represents the deviation from the perfect sp 2 hybridization of monolayer graphene as the result of defects/edges or the formation of chemical bonds between graphene -electrons and other molecules.
In fact, the intensifying (ID/IG) ratio in Figs For comparison, no discernible aging effects can be found in either the Raman spectra or XPS studies of undoped CVD-grown graphene stored in air, as exemplified in Fig. 4. These findings suggest that the reaction between the organic dopant and environmental moisture is the primary cause for the aging effect in doped graphene.
In Fig. 5(a) In general the most pronounced aging effect of doped monolayer graphene due to humidity is manifested by the increasing (ID/IG) intensity ratio, and the aging effect becomes mitigated in the intercalated bilayer graphene.
Next, we consider the dependence of the aging effect of both hole-and electron-doped graphene on the number of graphene layers (right panels of Fig. 5(a)-(l)). We found that all spectral characteristics exhibited much less time-dependent variations and therefore better stability for the intercalated bilayer graphene (blue symbols) than the doped monolayer graphene (red symbols), and the benefits of the intercalated bilayer configuration were more significant in the electron-doped samples.
To better illustrate the differences of the aging effect between hole-and electron-doped graphene, we compare in Fig. 6 Figure 7, the more significant time-dependent spectral degradation in electron-doped graphene suggested that the reaction of PEIE to moisture was more significant than that of PEDOT:PSS, which may be the result of a large number of (OH)  ions in PEIE ( Fig. 1(c)) that continuously reacted to environmental moisture.
We further investigated the XPS of all doped graphene samples aged under various conditions. As shown in Fig. 8(a)  Three specific molecular modes associated with water are indicated according to the spectrum of water (shown above (a)) taken from Reference [20].
In comparison with doped monolayer graphene in air, the aging effect on the XPS under 100% humidity appeared to be significantly worsened, as shown in Fig. 8 (c) C-1s and S-2p core electron spectra of hole-doped monolayer graphene in 100% humidity. Excess spectral features below the C-1s peak were found to increase with time. (d) C-1s and N-1s core electron spectra of electron-doped monolayer graphene in 100% humidity. Excess spectral features below the C-1s peak were also found to increase with time. (e) C-1s and S-2p core electron spectra of PEDOT:PSS intercalated bilayer graphene in air. (f) C-1s and N-1s core electron spectra of PEIE intercalated bilayer graphene in air. (g) SEM images of hole-doped monolayer graphene over a (4  4) m 2 area, from left to right: as-doped in air, after 30 days in air, and after 30 days in 100% humidity. (h) SEM images of electron-doped monolayer graphene over a (4  4) m 2 area, from left to right: as-doped in air, after 30 days in air, and after 30 days in 100% humidity.
To better understand the physical origin of excess electron doping under 100% humidity, we performed simulations based on the density functional theory (DFT) 23,24 to investigate the electronic interaction between carbon atoms in graphene and H2O molecules. As detailed in the supporting information, we find that when excess H2O molecules are placed close to the vicinity of graphene, they prefer to forming a monolayer that stretches parallel to the graphene sheet (Appendix and Figure 9), and there is a net electron transfer to the carbon atoms on graphene from the H2O molecules, with the hydrogen atoms of the water molecules being closer to the carbon atoms in graphene than the oxygen atoms (Appendix and Figure 10). Since carbon atoms with H2O molecules attached would be surrounded by more negative charges as compared to normal carbon atoms in graphene without H2O molecules, the electron binding energy of H2O-attached carbon atoms would be lower than that of the rest of carbon in graphene due to this physisorption process. As a result, there would be extra features appearing at the lower binding energy side of the XPS C-1s spectra, which is consistent with our experimental findings in

Figs. 8(c) and 8(d).
We further remark that this extra electron doping from saturated moisture to graphene is independent of the hygroscopic effects of the organic dopants that mediate chemical bonds to doped graphene, and is only significant under the 100% humidity condition.
In contrast, the aging effect on the XPS data of intercalated bilayer graphene appeared to be significantly mitigated relative to those of the monolayer graphene, as manifested in Fig. 8(e)-(f). For PEDOT:PSS intercalated bilayer graphene, both binding energies of the C-1s and S-2p core electrons appeared to be invariant up to 30 days, and the sp 2 hybridized C-OH and the sp 3 C-O components only exhibited slight increase relative to those of the hole-doped monolayer graphene ( Fig. 8(a)). In the case of PEIE intercalated bilayer graphene, both the binding energies of the C-1s and N-1s core electrons were downshifted by 0.1 eV in air after 30 days, which was smaller than the downshift of 0.2 eV for monolayer graphene in air after 30 days, suggesting that the influence of aging effects on the work function was reduced. Moreover, the increase of the C-C=O component with time was also weakened slightly for the intercalated bilayer graphene. These findings of weakened aging effects on intercalated bilayer graphene relative to doped monolayer graphene are consistent with those revealed from the Raman spectroscopic studies.
While the spectroscopic studies described above revealed the averaged effects of different conditions on the aging of graphene samples, additional investigation of spatially resolved SEM images of these graphene samples as a function of time can provide more information about the underlying mechanism of the aging effect. As exemplified in Fig. 8(g)-(h), both of the as-doped hole-and electrondoped monolayer graphene revealed relatively clean SEM images other than a few line defects, whereas dark spots appeared to develop along the defect lines and more defect lines proliferated after 30 days, particularly for samples stored in 100% humidity. We believe that the increasing dark spots and defect lines were associated with the development of degraded graphene, although we could not independently verify the local Raman spectra of the small dark spots and narrow lines due to the limited spatial resolution (~ 2m) of our Raman spectroscopy.

IV. DISCUSSION
Our spectroscopic studies of the aging effect of doped graphene suggest that the primary cause for degradation is due to the reaction of the hygroscopic organic dopants with environmental moisture, which results in reduction of carrier densities in the doped graphene and also creates excess edge states due to increasing damages propagated from the grain boundaries. The apparent difference in the severity of the aging effect between the hole-and electron-doped graphene may be attributed to the different numbers of reaction sites in the organic dopants rather than possible asymmetric behavior between the valence and conduction bands of graphene.
On the other hand, the improved spectral stability of doped graphene in the intercalated bilayer configuration suggests that the enclosure of organic dopants between two monolayers graphene can protect the organic molecules from fast environmental degradation. This finding differs from previous reports that primarily focused on the protective effect of stacking multiple graphene layers on top of the surface of the material of interest. 6 We believe that the configuration of intercalated bilayers can provide additional environmental protection because of the graphene layer between the intercalant and the underlying substrate: Our studies of all doped graphene samples were transferred to SiO2/Si substrates, which exhibited relatively rugged surfaces 25 and so could result in trapping of or pathways for environmental gases/liquids. Therefore, the insertion of a graphene monolayer at the interface of an organic dopant and the SiO2/Si substrate could significantly reduce the aging effect and improve the longterm stability of organic optoelectronic devices.

V. CONCLUSION
We have conducted systematic time-dependent Raman, XPS, UPS and SEM studies of organically doped graphene samples up to 30 days as a function of humidity, the number of graphene layers and the type of organic dopants. Detailed spectral analysis revealed that the aging effect of doped graphene was primarily associated with continuing reaction of organic dopants with environmental moister, which mediated excess chemical bonds to graphene and continuing reduction of doped carriers in graphene, and also induced proliferating damages along the grain boundaries of graphene. Additionally, high humidity can result in physisorption of a thin layer of water on the surface of graphene, giving rise to reduction in the electron binding energy of graphene. On the other hand, organic dopants in the intercalated bilayer configuration become better protected against environmental degradation. Therefore, low humidity, lowdefect large-area graphene sheets and the addition of an extra graphene layer at the interface between the organic dopant and the underlying substrate can provide significant long-term stability for organically doped graphene, lending better reliability for low-cost and flexible graphene-based electronic and optoelectronic devices.

ACKNOWLEDGMENTS
The work at Caltech was jointly supported by the National Science Foundation through the

MOLECULES ON GRAPHENE
We investigate how the presence of water molecules may influence the electronic structures of the carbon atoms in a graphene sheet by using a semi-empirical hybrid method known as the B3LYP method, 26 which was derived from the density functional theory (DFT). 23,24 The B3LYP method can provide high accuracy for molecular computation, and all optimization by the B3LYP method in this work was made by using a quantum-chemical calculation package, Gaussian 03 (Gaussian, Inc.), with a 6-31G(d,p) basis set.
For meaningful results attainable with a reasonable computation time, we use the ``coronene'' (see Figure 9) molecule to simulate reduced graphene. The optimized electronic structure of the system with the Mulliken charge 27 of each carbon atom specified is shown in Figure 9. Next, we add H2O molecules to the system and minimize the Gibbs free energy. We find that the Gibbs free energy is negative in the presence of H2O molecules and decreases with the increasing number of H2O molecules, which implies that it is energetically favorable for H2O molecules to react with graphene. The resulting Mulliken charge distributions for individual atoms in the assembly of reduced-graphene and 5 H2O molecules are shown in Figure 10(a). We note that the H2O molecules prefer to forming a monolayer stretching parallel to the reduced graphene rather than stacking above other H2O molecules, and the hydrogen molecules are closer to the graphene than the oxygen molecules, as exemplified in Figure 10(a) and 10(b) for the top view and side view, respectively, for a system of five H2O molecules and a coronene. The total Mulliken charge of H2O is negative, suggesting that H2O molecules lose electrons to carbon atoms, leading to lower binding energies for the C-1s spectrum, which is consistent with our XPS data.
FIG. 9. The Mulliken charges of carbon atoms in the "coronene", a piece of reduced graphene terminated by hydrogen atoms, from DFT calculations: The Mulliken charge for carbon atoms is represented in color scales. We note that the Mulliken charge for carbon atoms at equivalent positions is identical by symmetry.