-Near-Infrared Photoluminescence Spectral Imaging of Chemically Oxidized Graphene Flakes (cid:3)

In order to determine the local structure of carbon sp 2 clusters in chemically modiﬁed graphene oxide (GO) ﬂakes, their luminescence was analyzed, using near-infrared photoluminescence (NIR PL) spectral imaging. GO ﬂakes emit a broad PL spectrum of wavelengths of 800 to 1400 nm, indicating that they contain sp 2 clusters, whose size is theoretically estimated to be in the 1.3 to 2.3 nm region. The size distribution of such sp 2 clusters is fairly uniform at diﬀerent positions and for diﬀerent numbers of layers of GO, at the spatial resolution of NIR PL image (5 (cid:22) m). The analysis by transmission electron microscopy directly conﬁrmed the existence of such sp 2 clusters and the estimated size of the sp 2 clusters was widely ranged from 0.5 to 4 nm. The eﬀect of the GO reduction to the local structure of carbon sp 2 clusters was also studied and it was found that the NIR PL intensity decreased as the reduction progressed and that there was no large spectral shift. [DOI: 10.1380/ejssnt.2012.513]


I. INTRODUCTION
Graphene consists of a two-dimensional single atomic layer of sp 2 bonded carbon atoms and exhibits notable electronic and mechanical properties [1]. However, the applications of graphene based on its optical properties have been limited because defect-free graphene exhibits a zero band gap. Graphene oxide (GO) is a partially oxidized graphene sheet and has demonstrated interesting optical properties because we can introduce different gap behavior from that of graphene. GO contains nanometersize sp 2 carbon clusters isolated within an sp 3 carbon matrix [2,3]. Recently, the broadband photoluminescence (PL) of GO has been observed in the visible and nearinfrared (NIR) regions [4][5][6][7][8]. Although the exact mechanism of this PL emission remains unclear, many studies of PL in disordered carbon systems containing a mixture of sp 2 -and sp 3 -carbon have shown that the recombined localized electron-hole pairs within the sp 2 cluster domain behave as luminescence centers [4][5][6][7][8]. PL spectroscopy is a powerful tool for studying the local structure of sp 2 clusters in GO. As regards the size of the sp 2 clusters, theoretical calculations predict that larger clusters of diameter d (nm), should have a smaller energy gap, E, given by E ≈ 2/d (eV) [8]. For example, the energy gap for sp 2 clusters with d = 2 nm is 1 eV (∼1240 nm in wavelength). Thus, NIR PL is useful for studying nanometer-sized sp 2 clusters.
To understand the emission mechanism, PL behavior has been studied with thin films and liquid GO samples [7], and in terms of pH dependence [5], and reduction levels [6][7][8]. The reduction of GO makes it possible to change its optical and electronic properties. A common reduction method involves exposure to hydrazine vapor, which enables us to obtain an electrically conductive material [3]. As regards the change in the optical properties, a study of PL spectra obtained with UV and visible light showed that reducing GO weakens the PL intensity, but no spectral shift was observed [6]. The results do not indicate a significant change in the size of the clusters, and would suggest rather the formation of additional clusters.
In this study, we used NIR PL, transmission electron microscopy (TEM), and Raman spectroscopy to determine the size of the sp 2 clusters contained in an individual GO flake. We also studied the change in the structure of the sp 2 clusters induced by GO reduction.

A. Reagents and Materials
The natural graphite was a gift from Ito Kokuen Co., Inc. Hydrazine solution (35 wt% in water) and ammonia solution (28 wt% in water) were purchased from Aldrich and Kanto Chemical Co., Inc., respectively. Superior quality pure water was used in all the processes involved in aqueous solution preparation and washing.

B. Sample preparation
We synthesized an aqueous dispersion of GO from natural graphite using a modified Hummers method [3,9]. The GO dispersion was spin coated on the hydrophilic surface of a 10×10 (mm) quartz plate, prepared by treatments in piranha (H 2 O 2 :H 2 SO 4 =1:4) and NH 4 F. We prepared reduced GO by exposing the GO flakes on the plate to hydrazine vapor. A few µL of a mixture solution (hydrazine solution: ammonia solution = 1:7) was contained in the same vial with the plate, which was kept at 65 • C for 10 min (rGO-1) or at 95 • C for 60 min (rGO-2).

C. Characterization
Atomic force microscope (AFM) images were recorded in the AC (tapping) mode under ambient conditions using a D3100 Atomic Force Microscope (Digital Instruments). The samples for the AFM image were prepared by spincoating the GO dispersion on the hydrophilic surface of a quartz plate or a SiO 2 (285 nm)/Si plate, treated by the method described above.
The size of the sp 2 clusters was investigated by employing TEM with a JEOL ARM-200F at an acceleration voltage of 200 kV. Samples for the TEM images were prepared by pipetting a few µL of GO dispersion onto holey carbon mesh grids, which were then rinsed with distilled water. We used the reduction process described above to prepare rGO-2 on a mesh grid.

D. NIR PL and Raman measurement
Prior to the PL measurements, we obtained Raman images to determine the number of layers and the positions of the GO flakes on the plate. We measured the Raman and NIR PL spectra employing a Raman microprobe system (inVia Reflex/StreamLine microRaman spectrometer, Renishaw), and using a CCD and an InGaAs array detector for the Raman and PL measurements, respectively. The excitation light sources were the 532-and 785-nm lines of a laser-diode continuous-wave laser for the Raman and PL measurements, respectively. The maximum laser power focused on the sample with an objective lens (×50, NA 0.55) was about 35 mW (532 nm) or 7 mW (785 nm) for all the measurements. The spot size and the focal depth on the sample were about 1 µm and < 10 µm, respectively. The scattered light was collected by using a 180 • backscattering geometry, and the grating in the polychromator was 1800 line/mm (532 nm) or 600 line/mm (785 nm). The step sizes for collecting spectral images were < 2 µm and 5 µm for the Raman and PL measurements, respectively. We performed all the measurements at room temperature. Figure 1(a) shows an AFM image of the synthesized GO flakes deposited on a hydrophilic SiO 2 /Si surface. Most of the GO flakes were single-layer and had various shapes and sizes. The height of a single sheet was 1.3±0.1 nm ( Fig. 1(b)), similar to that reported in the literature [10]. This is significantly thicker than that of ideal graphene owing to the presence of oxygen-containing functional groups and adsorbed water above and below the carbon basal plane. Next, we characterized the GO flakes on the quartz plate by using an AFM image to determine their thickness, which means the number of layers in a flake ( Fig. 2(a)). The height of a single sheet was from 1.3 to 1.5 nm (Fig. 2(b)), which is similar to those observed in Fig. 1. The height at position A is the substrate level. The numbers of GO layers at positions B to D changed from single, to double, and triple layers. Then, we obtained a Raman image of the same GO flake (Fig. 2(c)). The Raman spectra extracted from the image at positions A to D show the typical spectral features of GO, namely it contains G and D bands at about 1590 and 1355 cm −1 , respectively [11]. The Raman intensities of both the G and D bands are roughly proportional to the number of layers. This means that we can determine the number of layers in any flake for further analysis simply from the Raman image. Figure 3(a) shows an NIR PL image of GO flakes at an emission wavelength of 1300 nm for several different numbers of GO layers, which were determined from a Raman image of the same GO flakes (Fig. 3(c)). The patterns agree well with the corresponding Raman image of the same GO flakes, indicating that the PL was derived from individual GO flakes. The PL spectra extracted from the image at positions A to C show a broad GO emission of 800 to 1400 nm ( Fig. 3(b)). According to theoretical cal- culations [8], the spectral range corresponds to the emission of sp 2 clusters whose d is from 1.3 to 2.3 nm. The Raman bands of the GO corresponding to the D and G bands were also observed around 874 and 894 nm in the PL spectra, which supports the view that the observed PL originated from GO flakes. The dip at around 1110 to 1130 nm is caused by the detector, not by the sample. Figure 3(d) shows the Raman spectra measured at positions A to C. The numbers of layers determined by the Raman intensity were A (triple), B (single), and C (none). Although the PL intensity increased with the number of GO layers, unlike the Raman intensity it was not proportional to the number of layers. This can be explained by the self quenching of the PL by the layered GO area, since GO has been reported to behave as a fluorescence quenching material [5,12]. Moreover, there was little spectral shift resulting from the position or the number of layers, indicating that the size distribution of the sp 2 clusters, which is sufficiently large for them to have small gaps that correspond to the NIR emission, is fairly uniform at different positions and for different numbers of layers, at least at this spatial resolution. A similar result has already been reported for GO luminescence in the UV-visible region [6]. Next, we studied the effect of reducing GO. We controlled the degree of reduction by controlling the reduction temperature and processing time. Figures 4(a) and

5(a)
show NIR PL images of rGO-1 and rGO-2 flakes, respectively, at an emission wavelength of 1300 nm, with several different numbers of layers, which were determined from the Raman images of the same rGO-1 and rGO-2 flakes (Figs. 4(c) and 5(c)). Again, the PL patterns agree well with the corresponding Raman images of the same rGO-1/rGO-2 flakes. There was no clear spectral shift caused by the reduction (Figs. 4(b) and 5(b)). Thus, it is difficult to determine whether a reduction induced a change in the size of the sp 2 clusters solely by using NIR PL.
We then conducted a TEM analysis to characterize the exact structures of the sp 2 clusters in the rGO-2 flakes (Fig. 6). Large transparent sheets of rGO-2 sample were observed on the grid (Fig. 6 (a)). The selected area electron diffraction pattern (SAD) of rGO-2 showed both diffraction rings and spots ( Fig. 6(b)). Compared with previously reported data for single-layer graphene [14,15] and chemically reduced graphene under hydrogen plasma [16], the crystal structure was not completely restored by hydrazine reduction. A high-resolution TEM (HRTEM) image clearly shows that the rGO-2 sheet consists of two different structures; graphitic areas and the amorphous sp 3 matrix, which was observed as fine and rough patterns, respectively (Fig. 6 (c)). The graphitic domains can be further separated into smaller sp 2 clus- ters. Figure 6(d) schematically shows the sp 2 clusters (indicated in pink) separated by linear defects (indicated by purple lines). We roughly estimated the size of the sp 2 clusters from the diameter of the inscribed circle in each sp 2 cluster. The clusters behave as luminescence centers and their diameter d determines the luminescence energy. The distribution of d showed that the size of the sp 2 clusters in the region varied from 0.5 to 4 nm (Fig. 7). The result is consistent with a previously reported TEM analysis of GO flakes synthesized with a similar method to ours, which reveals that a single GO flake contains sp 2 clusters with several different sizes ranging from 0.5 to 3 nm in the amorphous sp 3 matrix [16]. Thus, the distribution of the different size sp 2 clusters causes a broad NIR emission from GO in the 800 to 1400 nm range, whose size is theoretically estimated to be in the 1.3 to 2.3 nm region.
Raman spectroscopy is another useful technique with which to estimate the size of the sp 2 clusters in carbon materials such as graphene by using the Tuinstra-Koenig (TK) relation I D /I G = C(λ)/L a , where the crystallite size L a is obtained from x-ray data. Here, I D and I G represent the Raman intensities of the D and G bands, respectively, and C(λ) is a constant, which depends on the excitation laser wavelength λ [17]. However, for disordered carbon materials, which consist of smaller sp 2 clusters, namely where L a < 2 nm region, the TK re- lation is no longer valid and I D /I G ∝ L 2 a dependence works well [18]. By using I D /I G = 0.99 ( Fig. 2(d)) and 1.18 ( Fig. 4(d)) for GO and rGO-2, respectively, we found that the size of the sp 2 clusters increased by about 9% as a result of the reduction. Although the exact definitions of L a and d are different, the both are regarded as the size of the sp 2 clusters in our discussion. Together with the TEM analysis, the result indicates that the size distribution of the sp 2 clusters of GO is not largely different from that of rGO-2. It is therefore natural that no spectral shift was observed in the 800 to 1400 nm region when GO was reduced.
We also found that, as the reduction progressed, the PL intensity decreased (Figs. 4(b) and 5(b)). The decrease in the PL intensity is not likely caused by the change in the size distribution of the sp 2 clusters, according to our above Raman analysis. A possible reason is that the restoration of the line defects, which separate the graphitic domains into sp 2 clusters, is restored by hydrazine reduction. It may interrupt the recombination of localized electron-hole pairs within a sp 2 cluster and facilitate the relaxation of the excited states to the nonradiative recombination states [6]. However, the restoration mechanism of sp 2 clusters by reduction has not yet been directly confirmed. A further study will be needed to reveal the reason for the change in the PL intensity caused by the reduction of GO.

IV. CONCLUSIONS
In summary, GO flakes emit a broad NIR PL spectrum at wavelengths of 800 to 1400 nm, indicating that the GO flakes contain sp 2 clusters that are large enough to have a small energy gap that corresponds to the NIR emission. There is little spectral shift due to position or the number of layers, indicating that the size distribution of the sp 2 clusters is fairly uniform at least at the spatial resolution in our study. The TEM analysis of a reduced GO flake confirmed the existence of the sp 2 clusters. The size of the sp 2 clusters in rGO-2 flakes ranged from 0.5 to 4 nm, which can behave as luminescence centers in the NIR wavelength region. The cluster size was also evaluated by the Raman spectra, which showed that the size of the sp 2 clusters increased by about 9% as a result of the reduction. The little difference in the size distribution of the sp 2 clusters reasonably explains the little spectral shift in NIR PL caused by the reduction of GO.