STRUCTURAL INFORMATION FROM THE RAMAN SPECTRA OF ACTIVATED CARBON MATERIALS

Transcription

1 STRUCTURAL INFORMATION FROM THE RAMAN SPECTRA OF ACTIVATED CARBON MATERIALS N. Shimodaira, A.Masui, A.Takada, YShinozaki, and N. Tomita Research Center, Asahi Glass Co. Ltd Hazawa-cho, Kanagawa-ku, Yokohama, , Japan Introduction Raman spectrometry, is one of the most useful characterizing methods for carbon materials, because the spectral shape drastically changes not only due to the kind of abundant allotropic forms of carbon, but to the fine structural changes of the individual allotrope. For example, in polycrystalline graphites, two sharp peaks appear, G-band around 1580cm -1 and D-band around 1355 cm -1, which are generally ascribed to E 2g and A ig in-plane vibration modes, respectively. It is well known that the intensity ratio of D to G-band (I d /I g ) is correlated with the reciprocal of the crystallite size along basal plane (1/L a ) measured from XRD. [1-4] In sputtered or evaporated amorphous carbon films (a-c) regardless of hydrogenated or not, a relatively broad band around 1550 cm -1 overlapped with a broader band around 1400 cm -1 is observed. The Raman spectra are usefully deconvoluted into two peaks by using gaussian line shapes, which are also called G- and D-band in a-c. The variation in the position, the width, and the intensity ratio (I d /I g ) has been often examined as a function of deposition conditions and some properties measured from other techniques. Especially, the intensity ratio (I d /I g ) 1s used as the most useful parameter indicating the sp 2 cluster size, or the sp 3 to sp 2 bonding ratio in hydrogenated amorphous carbons. [ 5-11 ] As described above, the Raman spectra of crystalline graphites and a-cs have been studied in details so far, and the relationship between the spectral shape and the structure has been reported theoretically and qualitatively. However, in other non graphitic carbons such as activated carbons and carbon blacks, despite having the sp 2 bonding as a-c or graphite, the Raman spectra haven t been sufficiently understood relationally to the structure. In this paper, we report a novel characterizing method of the Raman spectra of activated carbon materials by curve fitting technique, and propose an interpretation on the microstructure. We also report the structural change before and after activation and the difference of the effect of activation on the structure between by steam and by alkali. Experimental Total 29 commercially available activated carbon powders were prepared. Among them, 16 samples were activated mainly by steam from non-graphitizable carbons, and 13 samples were activated mainly by alkali from graphitizable carbons. One of the former samples activated by steam from a carbonized phenol resin was annealed at various temperatures ranging from 1273 to 1673 K for 3 hours in argon. As references, three types of amorphous carbon films sputtered on silicon at different H 2 /Ar flow ratio at 6 mtorr, and 16 commercially available crystalline graphite powders, were prepared. All the a-c samples were also annealed at various temperatures ranging from 423 to 573 K for 1 hour in argon. In order to study the structural change before and after activation, a phenol resin and a pitch carbonized at 1173 K, were activated by steam at 1123 K and by KOH at 1023 K, respectively. Ramah spectra were obtained with Raman or T64000 (Jobin/ Yvon) using a nm Ar line as an excitation source. The laser was focused to about 100 µm in diameter at a power of less than 5 mw at the sample surface in order to prevent thermal degradation of the carbons. The spectra over the range from 800 to 2000 cm -1 at about 0.5 cm -1 intervals were measured in a backscattering geometry with a 64 cm single monochromater, and detected by 16 bit CCD camera with 1024ch. Because of the low signal to noise, 5 to 10 scans were averaged. Fitting analysis with four Gaussian curves on a linear background was conducted on all the Raman spectra in the region from 800 to 2000 cm -1.Raman spectra of a-cs were also deconvoluted to G and D-bands by using two Gaussians and a straight line as the background. X-ray diffraction was performed with a Rigaku RINT2500 with CuKα radiation at 40 kv and 200 ma. Microstructures were observed with a high resolution transmission electron microscope (HITACHI H9000) operating at 300 kv. Results and Discussion The examples of Raman spectra with fitting results of the activated carbon samples are represented in Figure 1, (a) for a non-graphitizable carbon and (b) for a graphitizable carbon, respectively. Note that this fit using four Gaussians faithfully reproduces the experimental data, and that two rela-

2 tively sharp peaks at about 1600 cm -1 and about 1350 cm -1, and two relatively broad peaks around 1560 cm -1 and around 1340 cm -1, namely G1, D1, G2 and D2, are observed in both the samples. All Raman spectra examined in this study showed excellent curve-fit (accuracy factor ~0.99) without exception. The behavior of each deconvoluted peak in all the samples becomes clear by plotting the Raman shift and the full width at half maximum (FWHM) in two dimensionally, as shown in Figure 2. Focusing on the annealed samples (closed circle), each peak regularly moves on the 2D-plot space with the increase of temperature as indicated by arrows. It is quite notable that the plot points of each peak in all the activated carbon samples (closed square) concentrate on almost the same regions as those in the annealed samples, though the points for D2 peak are widely scattered. In Fig.2, the fit results of a-c films (plus) and polycrystalline graphites (x) are also shown. All the positions of G1 and D1 peaks are almost constant and close to those of G and D-band of the graphites, respectively. On the other hand, G2 and D2 points disperse in the close regions to G and D- band of a-cs, respectively, and the upward shift of the position and the decrease of the FWHM in G2 peak, as anneal temperature increases, is in accord with that in G-band of a- C. This trend is similar to the observation in a-c films reported by others [5]. These features give the view that the activated carbon materials may be composed of both the graphite crystal and amorphous structure, and that from each component G-and D-band may appear as G1 and D1 peaks, and G2 and D2 peaks, respectively. However, XRO diffraction patterns showed no sharp line in any samples (Fig.3). High resolution TEM microstructures showed no crystal lattice image (Fig.4), and only winding (002) basal planes (black contrast), which are extremely unclear in a-c, can be homogeneously observed in the samples. It is apparent that the crystal grain structure like graphite does not exist in the activated carbon materials. We propose that the basic structure of the activated carbon materials consists of six-fold aromatic rings having sp 2 bondings and includes both bond-angle disorder networks as given in Beeman s model [12], and order networks, and that the G and D-band from each the honeycomb network structure are exclusively observed in the Raman spectra. Because of the positions, G1 and D1 peaks are respectively ascribed to E 2g and A lg in-plane vibration modes from the order structure in the sp 2 clusters, not from crystal structure like graphite. On the other hand, the bond-angle disorder structure behaves like a-c and is the origin of G2 and D2 peaks. The upward shift of G2 peak and the decrease of FWHM of G2 and D2 peaks, as anneal temperature increases, are due to the removal of bond-angle disorder, as indicated by R.O.Oillon in a-c film [5]. The peak intensity ratios, ID1/IG1, ID2/IG2, and IG2/IG1, are plotted as a function of anneal temperature in Fig. 5. The decrease of IG2/IG1 with increasing temperature indicates the decline of disorder structure due to thermal relaxation. Other peak ratios, ID1/IG1 and ID2/IG2, don t clearly show the dependence on anneal temperature, probably because the structural transition from disorder to order complicate the size effect of each structure on the Raman spectral shape. Consequently, it is possible that the Raman spectral shape of activated carbon materials sensitively changes with the content of the disorder honeycomb structure (IG2/IG1) and the degree of the thermal relaxation. Since the G2 peak position is the most convenient parameter expressing the degree of the relaxation, the peak ratios in all the activated carbon samples are plotted against it in Fig. 6. It is reasonable that the trend of the decrease of IG2/IG1 with the increase of the G2 peak position is observed. On the whole, the ID2/ IG2 tends to increase with the increase of the G2 peak position as indicated in a-c films [5, 8, 9], while the ID1/IG1 doesn t change so much around 1.0. This suggests that the variety of the Raman spectral shape in activated carbon materials is mainly dominated by the disorder structure. In Fig. 7, the fitted Raman curves before and after activation are represented, (a) for a carbonized phenol resin activated by steam and (b) for a carbonized pitch activated by KOH. The deconvoluted peak parameters including the peak ratios are given in Table. It is worth noting that the change of the peak parameters as well as the spectral shape accompanied by activation proceeds just in opposite direction between by steam and by KOH. That is, it is concluded on the basis of the above interpretation that the activation by steam has a removal effect of the disorder structure, while the activation by KOH induces the disorder structure. Conclusion The structure of activated carbon materials is characterized by using oaussian peaks, G1, G2, D1, and D2, fitted for Raman spectra. In 20-plot of the Raman shift vs. FWHM, the distributed regions of each peak are mutually exclusive, and the G1 position is stable to heat treatment, while G2, D1, and D2 peaks are very sensitive. From the behavior of these peaks and the results of XRD and TEM observation, it is suggested that G1 and D1 peaks are due to the order structure of honeycomb networks, while G2 and D2 peaks are due to the disorder structure with bond-angle distortion in the networks. The peak ratio, IG2/IG1, can be used as a useful parameter expressing the content of disorder structure. The structural change before and after activation was also exam-

3 ined by this technique. It is concluded on the basis of the above interpretation that the activation by steam has a removal effect of the disorder structure, while the activation by KOH induces the disorder structure. References [1] F.Tuinstraand J.L. Koenig, J. Chem. Phys. 53, 1126 (1970) [2] G. Katagiri, H. Ishida, and A. Ishitani, Carbon 26, 565 (1988). [3] P. Lespade, A. Marchand, M.Couzi, and F. Cruege, Carbon 22, 375 (1984). [4] Y. Wang, D. C.AIsmeyer, and R. L. McCreery, Chem. Mater. 2,557 (1990). [5] R. 0. Dillon, J, A. Woollam, and V. Katkanant, Phys. Rev. B 29,3482 (1984). [6] M. Ramsteiner and J. Wagner, Appl. Phys. Lett. 51, 1355 (1987). [7] M. Yoshikawa, G. Katagiri, H. Ishida, A. Ishitani, and T. Akamatsu, J. Appl. Phys. 64, 6464 ( 1988). [8] N-H. Cho, K. M. Krishnan, D. K. Veirs, M. D. Rubin, C. B. Hopper, B. Bhushan, and D. B. Bogy, J. Mater. Res. 5,2543 (1990). [9] J. W. Ager III, IEEE Trans. Magn. 29, 259 (1993). [10] M. A. Tamor and W. C. Vassell, J. Appl. Phys. 76, 3823 (1994). [11] A. Cuesta, P. Dhamelincourt, J. Laureyns, A. Martinez- Alonso, and J. M. D. Tascon, Carbon 32,1523 (1994). [12] D. Beeman, J. Silverman, R. Lynds, and M. R. Anderson, Phys. Rev. B 30, 870 (1984).

4 Figure 1. Raman spectra with fitting results; (a) a non-graphitizable carbon activated by steam, (b) a graphitizable carbon activated by alkali. Figure 2. 2D-Plots of decomvoluted peak parameters: peak position vs. FWHM. Figure 3. XRD patterns of the activated carbons shown in Fig. 1.

5 Figure 4. High resolution TEM images of the activated carbons shown in Fig. 1. Figure 5. Peak ratios as a function of anneal temperature; (a) ID1/IG1, (b) ID2/IG2, and (c) IG2/IG1. Figure 6. Peak ratios as a function of G2 peak position; ID1/IG1, (b) ID2/IG2, and (c) IG2/IG1. Figure 7. Raman spectra (fitted curve) before and after activation; (a) by steam and (b)by KOH.

6 Table. Fitting parameters (v: positions, W: widths) of the Raman spectra in Fig.7.