- Email: [email protected]
Carbons and Graphites: Surface Properties of
Carbons and Graphites: Surface Properties of The surface properties of carbons and graphites are critical to their performance in most of their practical applications. There are three principal forms of carbon, namely diamond, various fullerenes, and graphite. The surface properties of these materials will be dependent upon surface chemical features and physical topography. Carbons and graphites can be prepared with a very wide range of surface areas as the surface topography is adjusted. In addition, the behavior of the surface can be altered by the arrangement of carbon atoms in the surface region, and by changing the surface chemistry to make the carbon surface capable of greater chemical reactivity. Fullerenes are distinct molecules, but can be used as materials when they are combined together to make buckey tubes and similar materials. A full discussion of these materials is given elsewhere (see Fullerene Formation). Diamond (see Diamond: Natural) has an important surface chemistry. An excellent review of diamond surface chemistry and a comprehensive biography has been published by Wei and Yates (1995). The surface chemistry of diamond, being a three-dimensional structure where all the carbon atoms are in a tetrahedral environment (consistent with a model where the carbon atoms are sp$ hybridized) depends upon how the surface crystal plane is generated at the surface, and thus what ‘‘dangling’’ bonds (from half empty orbitals) are produced. For example, a (100) diamond surface has two ‘‘dangling’’ bonds, while a (111) and (110) surface have only one ‘‘dangling’’ bond. Normally the diamond surface reconstructs to minimize the number of ‘‘dangling’’ bonds. Chemisorption can occur on the diamond surface, and the ability of this surface to grow through processes such as chemical vapor deposition has meant that an understanding of the surface chemistry of diamond is important in understanding the growth of diamond films. Graphite, having a two-dimensional structure (consistent with a model where the carbon atoms are sp# hybridized), has a planar surface that results when graphite is cleaved. The bonding within the two-dimensional sheets is strong, while the graphite sheets are only weakly held together. The natural cleavage direction lies along the graphite sheets, and the reactivity of this surface is very site-dependent, the edge sites being the sites of chemical reactivity. There are an enormous number of carbons whose structure is based upon that of graphite, but where the local order is very short range (i.e., the graphite sheets are relatively small, and arranged in an overall order that can be very random). In these materials there can be considerable variations in the number of edge and basal plane sites on the surface, and this will have a major impact on their chemical reactivity. Carbon materials based upon graphite are very important as
absorbent materials, and their ability to perform in this way will depend upon the topography of the surface (e.g., the presence or absence of small pits and slits that can adsorb molecules by physical adsorption), and the number of chemically active sites. Carbon fibers, where the surface area is low (as opposed to high surface area activated fibers), but with a surface that has been subjected to surface oxidation, form the essential strengthening element in composite materials in many practical applications. The internal surfaces between the carbon fibers and the matrix material of these composites are central to the mechanical properties of the final composite. 1. Probing the Surface Chemistry and Topography of Carbon and Graphites The surface topography of carbon and graphites can be determined by traditional methods such as transmission electron microscopy (TEM), scanning electron microscopy (SEM), scanning tunneling microscopy (STM), atomic force microscopy (AFM), and a variety of topographically sensitive surface analytical methods. In the case of single crystal diamond, methods such as low-energy electron diffraction and angle-resolved photoelectron spectroscopy provide valuable information, but in the case of disordered carbons based upon graphite these methods are less effective. Most surface analytical methods depend upon subjecting the surface to an electron, ion, neutral atom, or photon beam, and observing the presence of these entities from the surface. All these methods may cause damage to the surface, so that the surface probed is the damaged surface rather than the real surface. Surface scientists take care to identify any surface damage, and surface analysis combined with a good predictive theoretical model that can effectively predict the surface analysis result from a defined system provides an essential test for surface damage. Graphite is very sensitive to damage from ions and electrons, which can lead to break-up of the graphitic sheets. There are a range of surface analytical techniques that can be applied to the study of carbons and graphites. Some of these are inherently surface sensitive, such as x-ray photoelectron spectroscopy (XPS), Auger spectroscopy, secondary-ion mass spectroscopy (SIMS), high-resolution electron energy loss spectroscopy (HREELS), and ion scattering spectroscopy, while others may provide surface analytical information in appropriate cases, such as x-ray diffraction (XRD), infrared spectroscopy, and Raman spectroscopy. 1.1 X-ray Diffraction XRD has been used extensively for the examination of materials and thin films. Its effective use depends upon 1
Carbons and Graphites: Surface Properties of
Figure 1 AFM image of a 5i5 µm# area of a high-modulus pitch-based carbon fiber coated with a film of silicon carbide (after Rooke and Sherwood 1994).
having a crystalline material. The technique is a bulksensitive analytical method, but can be used to provide information relevant to surface changes in suitable circumstances. For example, XRD can provide useful information on the extent to which surface treatment of a carbon system has affected the bulk of the material. It is possible to use XRD in a thin-film mode, employing very small take-off angles, to derive some surface information, but generally speaking it must be regarded as a bulk structural technique. Examples of the application of XRD to the study of carbon fibers can be found in the author’s own work (Xie and Sherwood 1990a, 1990b, 1991a, 1991b, 1992).
Raman spectroscopy is a highly effective method for distinguishing between diamond and graphite, the former giving a narrow peak at 1332 cmV" and the latter two wide peaks at 1345 cmV" and 1540 cmV" with a scattering cross-section some 50 times greater than that of diamond. This high sensitivity for graphite can be employed as a useful tool to monitor the quality of thin diamond films. The sampling depth of FTIR and Raman spectroscopies is considerable, which means that if one is attempting to identify surface functionality and there is chemical functionality in the interior of the carbon, then the information is ambiguous.
1.2 Fourier Transform Infrared and Raman Spectroscopies
1.3 High-resolution Electron Energy Loss Spectroscopy
Fourier transform infrared (FTIR) and Raman spectroscopies can in principle provide information about surface functionality on carbons and graphites.
HREELS provides the ability to provide surface sensitive vibrational information. For example, the presence of C-H and C-D sites on diamond is readily
2
Carbons and Graphites: Surface Properties of determined by this approach (Thoms et al. 1994). On graphite systems, care must be taken to avoid sample damage from the electron beam (discussed below).
1.4 Scanning Electron Microscopy SEM has been so widely used for topographical studies of carbons that it will not be discussed here. The use of electron beams can lead to substantial sample damage, and the SEM study itself is likely to remove any surface functionality from the carbon.
1.5 Scanning Tunneling Microscopy and Atomic Force Microscopy Both of these approaches have proved very valuable for probing carbon surface topography. For example, Fig. 1 shows the AFM image of a carbon fiber coated with a surface film of silicon carbide (Rooke and Sherwood 1994).
1.6 Secondary-ion Mass Spectrometry SIMS is a highly surface sensitive method, but one which can also cause sample damage because of the fragile nature of carbons. Time-of-flight SIMS has proved valuable for a number of carbons, including carbon fibers.
1.7 Other Approaches Other approaches can be found discussed elsewhere (e.g., Sherwood 1998).
2. Photoelectron Spectroscopic Studies The surface chemistry of carbon is not well suited to many surface analytical probes. Probes that are not inherently surface sensitive cannot reliably be used because functionalized carbon lying within the bulk of a carbon cannot be separated from functionalized carbon lying at the surface. The least potentially destructive surface analytical probe that provides considerable chemical information is photoelectron spectroscopy. Core XPS has been used to probe the surface chemistry of a wide range of carbon compounds and carbon-containing polymers.
2.1 Carbon Spectra The photoelectron spectra of different types of carbon can be used to distinguish between the carbons.
(a) Core spectra. Core XPS shows a similar C1s region for the different types of carbon, but there are significant differences. The binding energy of all carbons lies at about 284.6 eV. In graphite this is accurately determined because of the conducting nature of graphite, thus the binding energy can be determined by the separation of the C1s peak and the Fermi level determined in the valence band spectra. In the case of diamond, an insulator, the binding energy is less easily determined. The band gap is 5.3 eV, and the separation between the C1s region and the top of the valence band indicates that the binding energy is comparable for diamond and graphite. The same is true of fullerenes. An important early paper discussed the details of the XPS spectra of carbon, with a focus on the valence band region (McFeely et al. 1974). The C1s of the conducting graphite and other graphitic carbons gives rise to an exponential tail that is seen on the core levels of conducting compounds. This tail, more correctly representable by a Doniach–SB unjic! lineshape (Doniach and SB unjic! 1970), arises from conduction band interaction associated with the movement of electrons from filled conduction band states to screen the positive charge formed in the photoelectron process. Diamond, an insulator, shows no such tail (McFeely et al. 1974), the same being true for fullerenes (Weaver 1992). The energy loss region of carbons (up to 60 eV higher binding energy than the C1s peak) shows characteristic features associated with the various loss processes, and was initially discussed by McFeely et al. (1974). This region continues to prove useful for determining chemical differences between different carbons (e.g., Yin et al. 1993). The energy loss region of fullerenes also gives characteristic features (e.g., Weaver 1992). The C1s peak in graphite and graphitic carbons has a linewidth (usually recorded as the full width at half maximum (FWHM)) that is sensitive to the extent of graphitic order (Viswanathan et al. 1997). Figure 2(a) shows how the width of this peak increases substantially after a carbon fiber is exposed to argon ion bombardment for 10 min. Small levels of oxidation are found on most carbons. For example, Fig. 2(b) shows the C1s region of an untreated carbon fiber. When monochromatic x radiation is used the improved resolution allows the small oxidation levels to be distinguished from the exponential tail. The fitting of this region, and the peak assignments will be discussed below. Some authors have suggested that there is a significant shift between the C1s from graphite (sometimes referred to as aromatic carbon) and the C1s from aliphatic carbon. In fact, there is little evidence that such a shift actually exists. For example, if one examines the C1s region of polystyrene (–CH CHPh–)n (where Ph represents a benzene ring) there #is no significant shift between the CH carbon (aliphatic carbon) and the carbon atoms# in the 3
Carbons and Graphites: Surface Properties of (a)
(b) Cls–PAN (M40) fiber
insulating hydrocarbon on conducting graphite) are in close proximity to one another.
Intensity (arb. units)
after 10 min etch
before etching
Intensity (arb. units)
achromatic
monochromatic FWHM (excluding tail) = 0.35 eV
290 285 Binding energy (eV) 293 285 277 Binding energy (eV)
Intensity (arb. units)
Figure 2 (a) Core XPS spectra of the C1s of a carbon fiber before and after exposure to argon ions. (b) Core XPS spectra of the C1s region of an untreated carbon fiber examined with and without monochromatic x radiation.
(a)
(b)
(c)
(d)
25
15
5 25 15 Binding energy (eV)
5
Figure 3 Valence band XPS spectra of four different types of carbon obtained with monochromatic Al Kα x radiation: (a) diamond; (b) crystalline graphite; (c) microcrystalline graphite; (d) glassy carbon (after McFeely et al. 1974).
benzene ring (aromatic carbon). Curve fitting of high-resolution data obtained with monochromatic radiation suggests a possible difference of only 0.2 eV (Beamson and Briggs 1992). Apparent shifts may occur, but these can easily arise from differential sample charging (e.g., see Dickinson et al. 1973) when carbon species of differing electrical conductivity (e.g., 4
(b) Valence band spectra. The valence band region of carbons shows some marked differences for different types of carbon. In the important early work of McFeely et al. (1974) an extensive study of carbon valence bands was conducted, and Fig. 3 shows the valence band region of graphite, diamond, and a glassy carbon from this work. This work pointed out that no difference would be expected in the valence band of a crystalline or a microcrystalline (polycrystalline) sample because the gross features of the density of states depend upon relatively shortrange order in the crystal. The principal difference in the valence band region appears between graphite and diamond. This difference can be understood when the spectra are compared with spectra generated by taking the density of states for different atomic orbitals from band structure calculations, and adjusting this by the photoelectron cross sections, and convoluting the result with the photon lineshape. Figure 4 compares these spectra for diamond and graphite. The calculations can be seen to reproduce the principal features of the experimental spectra, the vertical lines connecting the principal features in experimental spectra with the calculated spectra. The feature around 15 eV can be seen to be much more pronounced in the diamond spectra, and this difference is clearly predicted by the band structure calculations. Figure 10 shows the possible models that might be used for calculating carbon spectra. The reason for choosing coronene is that it is sufficiently large that it partially represents the graphite lattice but is sufficiently small that it can be functionalized to represent oxidized carbon (to be discussed below). The experimental valence band of benzene, well predicted by calculations (Riga et al. 1977, Sherwood 1992) shows sharp characteristic features. The calculated spectrum of coronene (Fig. 9(d)) shows a spectrum that is quite close to that of the graphite valence band, suggesting that the use of substituted coronenes is appropriate for predicting the valence band spectra of oxidized carbons.
2.2 Spectra of Oxidized Carbons XPS is particularly well suited to probe the surface chemistry of oxidized carbons. Oxidized carbons are important in many applications, including those where the surface of a carbon fiber is oxidized to enable better interaction between the fiber and the matrix into which the fiber is placed to make a composite material. The improved interaction arises because of improved
Carbons and Graphites: Surface Properties of
(a)
Calculated from band structure
20 10 0 Binding energy (eV)
Experimental Intensity (arb. units)
Intensity (arb. units)
Experimental
(b)
Calculated from band structure 30 20 10 0 Binding energy (eV)
Intensity (arb. units)
Figure 4 (a) Valence band XPS spectrum of graphite obtained with monochromatic Al Kα x radiation and the spectrum calculated using a band structure calculation for a twodimensional graphite lattice (after Viswanathan et al. 1997). (b) Valence band XPS spectrum of diamond obtained with monochromatic Al Kα x radiation and the spectrum calculated using a band structure calculation for a diamond lattice (the diamond experimental data is from McFeely et al. (1974), and had a nonlinear background removed; the calculation was carried out in a similar manner to the graphite calculation in Fig. 9 (P. M. A. Sherwood, unpublished results)).
290
0.5 V
2.0 V
1.0 V
2.5 V
1.5 V
3.0 V
285 290 Binding energy (eV)
285
Figure 5 Core XPS spectra of the C1s region of a polyacrylonitrilebased carbon fiber after potentiostatic oxidation to various potentials (shown versus the saturated calomel electrode) for 20 min in 2.7 mol dmV$ nitric acid solution. The spectra are fitted to six features at positions corresponding to the features in Table 1. A non-linear background is included in the fit (after Sherwood 1996a, 1996b).
wetting of the fiber surface by the matrix, and physical and chemical interaction between the oxidized fiber and the matrix. (For a more detailed discussion see, for example, Wang et al. (1998) and Sherwood (1996, 1998).) Core and valence band XPS provides valuable and complementary information that allows the functional groups on the oxidized surface to be identified, as well as an overall idea of the level of oxidation. Quantification in these systems is not possible except in the most well-defined cases. This is because quantification models in XPS require a number of assumptions whose use in carbon systems is often questionable. For example, quantification models are based upon assumptions such as a uniform homogeneous-composition flat layer of oxide on top of an elemental substrate. Except in the highest-quality graphite, carbons are rarely flat, and oxidation is rarely uniform. Variable take-off angle experiments, effective in flat systems, are of questionable value in many carbon systems such as fibers. The area ratio of different core peaks, adjusted for cross-section, electron escape depth, and the instrument transmission function can be a valuable measure for comparing the variation in the level of oxidation in a range of different samples. The C1s region shows a substantial shift for carbon\oxygen species, typical binding energies for such species being shown in Table 1. In the case of oxidized carbons it is usually necessary to use curvefitting methods to determine the positions and intensities of the overlapping peaks involved. Figure 5 shows a typical example of such a fitting approach for carbon fibers subjected to different levels of oxidation. It is important to remember that curve fitting approaches never give a unique solution. This means that it is essential to ensure that the fit is chemically reasonable, and to see if the same fitting approach can be applied to a system where the chemical composition changes in a systematic manner. In the case illustrated in Fig. 5 the level of oxidation can be seen to be increasing as a result of increasingly positive potentials to which the fibers are subjected. The fits in Fig. 5 were carried out so that the positions and widths of the fitted peak were consistent through the series of spectra. A further discussion of curve fitting in such systems can be found in a paper by Sherwood (1996b). The use of achromatic x radiation in carbon systems presents the problem that the oxidized carbon can be lost by surface decomposition in the spectrometer. This decomposition is largely caused by the heat from the target of the x-ray gun. This decomposition can be seen by substantial changes in the C1s region with increasing recording time, which are illustrated in Fig. 6 (Viswanathan et al. 1997). When monochromatic x radiation is used, where the x-ray gun is removed to a substantial distance from the sample, and high-energy Bremsstrahlung radiation is eliminated, no significant decomposition is seen. Comparison of the changes in the C1s and valence band region for different x-ray 5
Carbons and Graphites: Surface Properties of Monochromatic
(a)
Achromatic
(b)
0.2/1.5 h
Intensity (arb. units)
20 min
2h
Intensity (arb. units)
1.3/7 h
3/28 h
20 h
290
285
290
24/48 h
285
Binding energy (eV)
Figure 6 Core XPS spectra of the C1s region of a polyacrylonitrile based carbon fiber after galvanostatic oxidation in 1 mol dmV$ nitric acid solution at 0.5 A for 40 s studied using achromatic and monochromatic x-radiation for different exposure times (after Viswanathan et al. 1997).
Table 1 C1s features for different carbon\oxygen functionalities Peak No.
Functionality
Binding energy (eV )
Shift (eV )
1
Graphitic carbon
284.6
0
2
B carbon
285.3
0.7
3
C OH
286.1
1.5
4
Bridged structure
286.6–286.9
2.0–2.3
5
C O
287.6–288.2
3.0–3.6
6
COOH, COOR
288.8
4.2
7
CO#V $
289.1
4.5
8
π–π*
290.7
6.1
9
Plasmon
291.5
6.9
6
290
284 33 Binding energy (eV)
3
Figure 7 Core XPS spectra of (a) the C1s region and (b) the valence band XPS spectra of pitch-based carbon fibers after potentiostatic oxidation at 3.0 V (SCE) in 1 mol dmV$ nitric acid solution. Times of x-ray exposure are indicated on the right of the figure, the first number referring to the C1s region and the second number to the valence band region (after Xie and Sherwood 1990a, 1990b).
exposure times using achromatic radiation indicate that the C1s region shows the largest change in spectral appearance due to decomposition (Xie and Sherwood 1990b). The higher kinetic energy in the valence band region (about 1230 eV with Mg Kα x rays) compared to the C1s region (about 960 eV with Mg Kα x rays) would suggest that the latter region is more surface sensitive. However, initial inspection of the spectra can be misleading because the valence band region is composed of regions of significant oxygen and carbon intensity over a wide energy range. Thus, it is tempting to focus on the principal oxygen feature, the O2s region. The ratio of the O2s area (the peak around 25 eV in Fig. 7) to the total area in the valence band
Carbons and Graphites: Surface Properties of
12.34 8.67 7.04
10.58
6.74
8.92
4.00
8.85 5.85
2.59
8.77 5.45
2.27
Binding energy (eV) Figure 8 The calculated XPS valence band spectra of various substituted coronenes using multiple scattered wave Xα calculations (after Sherwood 1996a, 1996b).
region falls from about 0.32 to about 0.27 over the exposure times in Fig. 7 compared with the fall in the oxygen\carbon atomic ratio from the O1s\C1s peak areas from about 0.6 to about 0.4 for the same time. These differences will be discussed in more detail below, when it will be seen that the changes in the valence band region do correspond to significant decomposition, comparable to that seen in the C1s region. The valence band region provides complementary information to that found in the core region. Calculations for different functional groups on carbon using coronene with different types of functionality (Xie and Sherwood 1991a) provide a good prediction of what can be expected. Figure 8 shows that differences would be expected in the number of peaks in the principally O2s region around 25 eV (one for all the calculated functionalities except carboxyl which would be
expected to show two peaks), and in the region below 16 eV which shows a characteristic ‘‘fingerprint.’’ The position of the principally O2s region will be largely determined by the extent of interaction between O2s and C2s orbitals, which is different from the largely electrostatic effects that give rise to shifts in the core XPS regions. Thus, it is not surprising to find that the valence band region would behave differently than the core regions. This is predicted by the calculations, so while the C1s core region would be expected to give the same shift for hydroxide and epoxide functionalities, the valence band regions would be expected to be different, a prediction which is found to occur in spectra of oxidized carbon, where both of these species would be expected to be present (Xie et al. 1992). It is helpful to separate out the oxygen and carbon contributions to the valence band region. This has been done (P. M. A. Sherwood, unpublished results) in 7
Carbons and Graphites: Surface Properties of
(c)
(b)
(d)
Intensity (arb. units)
(a)
30
20
10
0 30 Binding energy (eV)
20
10
0
Figure 9 The calculated XPS valence band spectrum of a coronene and substituted coronene. The spectra in (a), (b) and (c) are for the ‘‘bridged’’ structure (Fig. 10), and the spectrum in (d) is for coronene. The substituted coronene spectrum for C O H #% (c) "# ' is shown in (a). Spectrum (b) shows the part of the spectrum in (a) corresponding to O2s and O2p intensity. Spectrum shows the part of the spectrum in (a) corresponding to C2s and C2p intensity. Spectra (b) and (c) are not adjusted for scale. The total area of spectrum (b) is 73% of the area of the total spectrum, and spectrum (c) is 27% of the area of the total spectrum.
Fig. 9 for the substituted coronene with the bridged structure (Fig. 10). The component peaks in the figure correspond to the position of all the energy levels in the valence band region with an area corresponding to the atomic orbital contribution adjusted by the photoelectron cross-section and a FWHM set at 2.58 eV. In fact, it is known that the widths of the component levels should vary within a spectrum due to effects such as vibrational broadening, but these are not easily predicted so all the peaks are set at the same width for predictive purposes. It is clear that the oxygen contribution (Fig. 9(b)) makes the principal contribution to the spectrum (73% of the intensity for C O H ). #%The"# heights ' of the spectra in Fig. 9 have all been adjusted so that the peak maxima are about the same for display purposes; if drawn to scale, Figs. 9(c) and 8
(d) would be of much lower intensity. It is also clear that the interaction between O2s and C2s electrons in the bridged structure leads to a considerable change in the distribution of carbon intensity when Fig. 9(c) (carbon contribution) is compared with the graphite ‘‘model’’ system coronene (C H ) in Fig. 9(d). Thus, the O2s region is referred to#% as"#being ‘‘principally’’ O2s above, because Fig. 9(c) indicates that significant C2s intensity moves into this region as a result of O2s–C2s interactions. There are also significant differences in the largely C2p region below 16 eV for Figs. 9(c) and (d), indicating that O2p–C2p interactions significantly impact spectral features. Hydrogen has such a small cross-section that it makes negligible contributions to all these spectra. Another important point that arises is that the principally O2s region around 25 eV binding
Carbons and Graphites: Surface Properties of
Benzene
Coronene Substituted coronene (bridged structure)
Figure 10 Models for calculating carbon spectra.
energy accounts for only 44% of the total oxygen intensity in the spectrum even though it is the most obvious feature in the spectrum. One can use the spectra in Fig. 9 to generate spectra for different carbon\oxygen atomic ratios. For example, Fig. 11 shows the valence band region for an oxidized carbon fiber obtained with monochromatic x radiation (which eliminates sample decomposition), for which the carbon\oxygen atomic ratio obtained from the C1s\O1s area ratios adjusted to give an atomic ratio as described above is 3. The calculated spectrum was obtained by first adjusting Fig. 9(b) to increase the width of peaks in the O2s region (the first seven energy levels of the 58 valence energy levels in the calculation) from a FWHM of 2.58 to 5.94 eV while keeping the peak areas the same (so keeping the contributions from each component energy level the same) to better reflect the observed widths of this region. Then the modified Fig. 9(b) was added to Fig. 9(c) and two times Fig. 9(d). This addition was performed in this way to ensure that one recognized that the C1s intensity in Fig. 9(c) results from carbon bonded to oxygen in the highly oxidized coronene shown in Fig. 10, and that unoxidized carbon is better represented by the coronene spectrum in Fig. 9(d). The agreement between the calculated and experimental spectra in Fig. 11 can be seen to be reasonable. The valence band calculation helps one reconcile the information provided in the core and the valence band regions. Figure 11 shows that when the core and valence band regions are recorded with monochromatic x radiation (when no decomposition is expected), then the carbon\oxygen atomic ratio is consistent for the core and valence band regions. This calculation helps us to reconcile the apparent observed
differences in the oxygen\carbon atomic ratio in the core region and the change in the O2s percentage in the valence band regions discussed above for the data in Fig. 7. When the ratio of the O2s area (the peak around 25 eV in Fig. 7) to the total area in the valence band region is adjusted to reflect the fact that some 66% of the oxygen intensity lies outside this region, and other helpful data from the calculation are included, an oxygen\carbon ratio can be obtained for the valence band region. The calculation indicates that for an oxygen\carbon ratio of 1, 81.5% of the valence band region will correspond to oxygen intensity, and 17% of the carbon intensity lies in the O2s region. Of course, one could obtain the first value from the photoelectric cross-section ratios (using cross-sections from Scofield (1976)), but such a calculation would require the assumption of a hybridization state for carbon and oxygen which would be necessarily more approximate than using the data from the calculation. An approximate measure of the oxygen\carbon ratio is obtained by taking the fraction of the area of the valence band region represented by the O2s region (let us call this x) and adjusting it by 1\0.44 to give the fraction of the valence band due to oxygen (2.27x). If we note that for an oxygen\carbon ratio of 1, about 3% of the O2s region is carbon intensity, then one can adjust this fraction to 2.35x for an average oxygen\ carbon ratio of 0.5 over this region. The difference between this value and the total area (1V2.35x) will represent the area due to carbon, which is then adjusted by (1\0.185) to reflect that fact that cross-section differences mean that when there is a oxygen\carbon ratio of 1, 81.5% of the valence band region will correspond to oxygen intensity. When these adjustments are performed, the oxygen\carbon 9
Carbons and Graphites: Surface Properties of
Intensity (arb. units)
Experiment
the substituted coronene of Fig. 10 linearly on the surface, then the thickness of oxidized carbon would be 1.28 AH (the C O bond length) and the oxygen\ carbon ratio from the valence band region should be 0.88 times that from the C1s region if the inelastic mean free path is 20 AH in the C1s region and 22.6 AH in the valence band region. Using these figures, a range of 0.62–0.37 from the C1s region should correspond to a range of 0.55–0.33 in the valence band region, very close to the observed figures listed above.
3. Concluding Comments
Calculated for C/O ratio = 3
Carbon presents some special challenges in the determination of its surface chemistry and topography, but the understanding of these properties is essential for the evaluation and understanding of many important properties of carbons. Core and valence band XPS techniques are probably the most valuable methods for the nondestructive evaluation of carbon surface chemistry, and AFM, STM, and electron microscopy are effective for the determination of surface topography.
Acknowledgments This material was based upon work supported by the National Science Foundation (under Grant No. CHE9421068), NASA, and the State of Kansas. 31
23
15
7
Binding energy (eV)
Figure 11 Comparison of the calculated XPS valence band spectrum of polyacrylonitrile-based carbon fibers after galvanostatic oxidation in 1 mol dmV$ nitric acid solution at 0.25 A for 40 s (experimental data from Viswanathan et al. 1997). The calculated spectrum was performed by combining the data in such a way that the carbon\oxygen ratio was 3 (in agreement with the experimental evaluation based on the C1s to O1s intensity ratio). The way in which the calculated spectrum was generated is explained in the text.
ratio in the valence band region of Fig. 7 ranges from 0.56 to 0.32. The change in the oxygen\carbon ratio calculated in this way is thus more striking than the change in the percentage of the valence band in the O2s region (0.32–0.27), indicating how a visual inspection of the spectral features can be misleading. This calculated atomic ratio is in line with the oxygen\carbon ratio range from the C1s and O1s region, namely 0.62–0.37. The greater surface sensitivity of the C1s region would lead one to expect a greater oxygen\carbon ratio using C1s rather than valence band data. For example, if one had a flat carbon surface that consisted of the functionality of 10
Bibliography Beamson G, Briggs D 1992 High Resolution XPS of Organic Polymers—The Scienta ESCA300 Database. Wiley, Chichester, UK Dickinson T, Povey A F, Sherwood P M A 1973 Differential sample charging in ESCA. J. Electron Spectrosc. Relat. Phenom. 2, 441–7 Doniach S, SB unjic! 1970 Many-electron singularity in x-ray photoemission and x-ray line spectra from metals. J. Phys. C 3, 285–91 McFeely F R, Kowalczyk S P, Ley L, Cavell R G, Pollak R A, Shirley D A 1974 X-ray photoemission studies of diamond, graphite, and glassy carbon valence bands. Phys. Re. B 12, 5268–78 Riga J, Pireaux J J, Verbist J J 1977 An ESCA study of solid benzene. Mol. Phys. 34, 131–8 Rooke M A, Sherwood P M A 1994 x-ray photoelectron spectroscopic study of the effect of ion etching of silicon carbide on a carbon fiber. Carbon 33, 375–80 Scofield J H 1976 Hartree–Slater subshell photoionization crosssections at 1254 and 1487 eV. J. Electron Spectrosc. Relat. Phenom. 8, 129–37 Sherwood P M A 1992 Carbon, functionalized carbon, and hydrocarbons studied by valence band photoemission. J. Vac. Sci. Technol. A 10, 2783–7 Sherwood P M A 1996a Surface analysis of carbon and carbon fibers for composites. J. Electron Spectrosc. 81, 319–42 Sherwood P M A 1996b Curve fitting in surface analysis and the
Carbons and Graphites: Surface Properties of effect of background inclusion in the fitting process. J. Vac. Sci. Technol. A 14, 1424–32 Sherwood P M A 1998 Composites. In: Rivie' re J C, Myhra S (eds.) Handbook of Surface and Interface Analysis. Dekker, New York, pp. 605–41 Thoms B D, Russel J N Jr., Pehrsson P E, Butler J E 1994 Adsorption and abstraction of hydrogen on polycrystalline diamond. J. Chem. Phys. 100, 8425–31 Viswanathan H, Rooke M A, Sherwood P M A 1997 X-ray photoelectron spectroscopic studies of carbon fiber surfaces. 21 Comparison of carbon fibers electrochemically oxidized in acid using achromic and monochromatic XPS. Surface Interface Anal. 25, 409–17 Wang T K, Donnet J-B, Peng J C M, Rebouillat S 1998 In: Donnet J-B, Wang T K, Peng J C M, Rebouillat S (eds.) Carbon Fibers, 3rd edn. Dekker, New York, pp. 231–309 Weaver J H 1992 Fullerenes and fullerides: photoemission and scanning tunneling microscopy studies. Acc. Chem. Res. 25, 143–9 Wei J, Yates J T Jr. 1995 Diamond surface chemistry I—a review. Diamond surface chemistry II—selected bibliography. Crit. Re. Surface Chem. 5, 1–248 Xie Y, Sherwood P M A 1990a x-ray photoelectron spectroscopic studies of carbon fiber surfaces. Part XII: the effect of
microwave plasma treatment on pitch-based carbon fiber surfaces. Appl. Spectrosc. 44, 797–803 Xie Y, Sherwood P M A 1990b X-ray photoelectron spectroscopic studies of carbon fiber surfaces. Part XIV: electrochemical treatment of pitch-based fibers and the surface and bulk structure changes monitored by XPS, XRD, and SEM. Appl. Spectrosc. 44, 1621–8 Xie Y, Sherwood P M A 1991a X-ray photoelectron spectroscopic studies of carbon fiber surfaces. 13. Valence-band studies of oxidized fibers interpreted by xα calculations. Chem. Mater. 3, 164–8 Xie Y, Sherwood P M A 1991b X-ray photoelectron spectroscopic studies of carbon fiber surfaces. Part XV: electrochemical treatment on pitch-based fibers by potentiostatic and galvanostatic methods. Appl. Spectrosc. 45, 1158–65 Xie Y, Wang T, Franklin O, Sherwood P M A 1992 X-ray photoelectron spectroscopic studies of carbon fiber surfaces. Part XVI: core-level and valence-band studies of pitch-based fibers electrochemically treated in ammonium carbonate solution. Appl. Spectrosc. 46, 645–51 Yin M, Koutsky J A, Barr T L, Rodriguez N M, Baker R T K, Klebanov L 1993 Characterization of carbon microfibers as reinforcement for epoxy resins. Chem. Mater. 5, 1024–31
P. M. A. Sherwood
Copyright ' 2001 Elsevier Science Ltd. All rights reserved. No part of this publication may be reproduced, stored in any retrieval system or transmitted in any form or by any means : electronic, electrostatic, magnetic tape, mechanical, photocopying, recording or otherwise, without permission in writing from the publishers. Encyclopedia of Materials : Science and Technology ISBN: 0-08-0431526 pp. 985–995 11
absorbent materials, and their ability to perform in this way will depend upon the topography of the surface (e.g., the presence or absence of small pits and slits that can adsorb molecules by physical adsorption), and the number of chemically active sites. Carbon fibers, where the surface area is low (as opposed to high surface area activated fibers), but with a surface that has been subjected to surface oxidation, form the essential strengthening element in composite materials in many practical applications. The internal surfaces between the carbon fibers and the matrix material of these composites are central to the mechanical properties of the final composite. 1. Probing the Surface Chemistry and Topography of Carbon and Graphites The surface topography of carbon and graphites can be determined by traditional methods such as transmission electron microscopy (TEM), scanning electron microscopy (SEM), scanning tunneling microscopy (STM), atomic force microscopy (AFM), and a variety of topographically sensitive surface analytical methods. In the case of single crystal diamond, methods such as low-energy electron diffraction and angle-resolved photoelectron spectroscopy provide valuable information, but in the case of disordered carbons based upon graphite these methods are less effective. Most surface analytical methods depend upon subjecting the surface to an electron, ion, neutral atom, or photon beam, and observing the presence of these entities from the surface. All these methods may cause damage to the surface, so that the surface probed is the damaged surface rather than the real surface. Surface scientists take care to identify any surface damage, and surface analysis combined with a good predictive theoretical model that can effectively predict the surface analysis result from a defined system provides an essential test for surface damage. Graphite is very sensitive to damage from ions and electrons, which can lead to break-up of the graphitic sheets. There are a range of surface analytical techniques that can be applied to the study of carbons and graphites. Some of these are inherently surface sensitive, such as x-ray photoelectron spectroscopy (XPS), Auger spectroscopy, secondary-ion mass spectroscopy (SIMS), high-resolution electron energy loss spectroscopy (HREELS), and ion scattering spectroscopy, while others may provide surface analytical information in appropriate cases, such as x-ray diffraction (XRD), infrared spectroscopy, and Raman spectroscopy. 1.1 X-ray Diffraction XRD has been used extensively for the examination of materials and thin films. Its effective use depends upon 1
Carbons and Graphites: Surface Properties of
Figure 1 AFM image of a 5i5 µm# area of a high-modulus pitch-based carbon fiber coated with a film of silicon carbide (after Rooke and Sherwood 1994).
having a crystalline material. The technique is a bulksensitive analytical method, but can be used to provide information relevant to surface changes in suitable circumstances. For example, XRD can provide useful information on the extent to which surface treatment of a carbon system has affected the bulk of the material. It is possible to use XRD in a thin-film mode, employing very small take-off angles, to derive some surface information, but generally speaking it must be regarded as a bulk structural technique. Examples of the application of XRD to the study of carbon fibers can be found in the author’s own work (Xie and Sherwood 1990a, 1990b, 1991a, 1991b, 1992).
Raman spectroscopy is a highly effective method for distinguishing between diamond and graphite, the former giving a narrow peak at 1332 cmV" and the latter two wide peaks at 1345 cmV" and 1540 cmV" with a scattering cross-section some 50 times greater than that of diamond. This high sensitivity for graphite can be employed as a useful tool to monitor the quality of thin diamond films. The sampling depth of FTIR and Raman spectroscopies is considerable, which means that if one is attempting to identify surface functionality and there is chemical functionality in the interior of the carbon, then the information is ambiguous.
1.2 Fourier Transform Infrared and Raman Spectroscopies
1.3 High-resolution Electron Energy Loss Spectroscopy
Fourier transform infrared (FTIR) and Raman spectroscopies can in principle provide information about surface functionality on carbons and graphites.
HREELS provides the ability to provide surface sensitive vibrational information. For example, the presence of C-H and C-D sites on diamond is readily
2
Carbons and Graphites: Surface Properties of determined by this approach (Thoms et al. 1994). On graphite systems, care must be taken to avoid sample damage from the electron beam (discussed below).
1.4 Scanning Electron Microscopy SEM has been so widely used for topographical studies of carbons that it will not be discussed here. The use of electron beams can lead to substantial sample damage, and the SEM study itself is likely to remove any surface functionality from the carbon.
1.5 Scanning Tunneling Microscopy and Atomic Force Microscopy Both of these approaches have proved very valuable for probing carbon surface topography. For example, Fig. 1 shows the AFM image of a carbon fiber coated with a surface film of silicon carbide (Rooke and Sherwood 1994).
1.6 Secondary-ion Mass Spectrometry SIMS is a highly surface sensitive method, but one which can also cause sample damage because of the fragile nature of carbons. Time-of-flight SIMS has proved valuable for a number of carbons, including carbon fibers.
1.7 Other Approaches Other approaches can be found discussed elsewhere (e.g., Sherwood 1998).
2. Photoelectron Spectroscopic Studies The surface chemistry of carbon is not well suited to many surface analytical probes. Probes that are not inherently surface sensitive cannot reliably be used because functionalized carbon lying within the bulk of a carbon cannot be separated from functionalized carbon lying at the surface. The least potentially destructive surface analytical probe that provides considerable chemical information is photoelectron spectroscopy. Core XPS has been used to probe the surface chemistry of a wide range of carbon compounds and carbon-containing polymers.
2.1 Carbon Spectra The photoelectron spectra of different types of carbon can be used to distinguish between the carbons.
(a) Core spectra. Core XPS shows a similar C1s region for the different types of carbon, but there are significant differences. The binding energy of all carbons lies at about 284.6 eV. In graphite this is accurately determined because of the conducting nature of graphite, thus the binding energy can be determined by the separation of the C1s peak and the Fermi level determined in the valence band spectra. In the case of diamond, an insulator, the binding energy is less easily determined. The band gap is 5.3 eV, and the separation between the C1s region and the top of the valence band indicates that the binding energy is comparable for diamond and graphite. The same is true of fullerenes. An important early paper discussed the details of the XPS spectra of carbon, with a focus on the valence band region (McFeely et al. 1974). The C1s of the conducting graphite and other graphitic carbons gives rise to an exponential tail that is seen on the core levels of conducting compounds. This tail, more correctly representable by a Doniach–SB unjic! lineshape (Doniach and SB unjic! 1970), arises from conduction band interaction associated with the movement of electrons from filled conduction band states to screen the positive charge formed in the photoelectron process. Diamond, an insulator, shows no such tail (McFeely et al. 1974), the same being true for fullerenes (Weaver 1992). The energy loss region of carbons (up to 60 eV higher binding energy than the C1s peak) shows characteristic features associated with the various loss processes, and was initially discussed by McFeely et al. (1974). This region continues to prove useful for determining chemical differences between different carbons (e.g., Yin et al. 1993). The energy loss region of fullerenes also gives characteristic features (e.g., Weaver 1992). The C1s peak in graphite and graphitic carbons has a linewidth (usually recorded as the full width at half maximum (FWHM)) that is sensitive to the extent of graphitic order (Viswanathan et al. 1997). Figure 2(a) shows how the width of this peak increases substantially after a carbon fiber is exposed to argon ion bombardment for 10 min. Small levels of oxidation are found on most carbons. For example, Fig. 2(b) shows the C1s region of an untreated carbon fiber. When monochromatic x radiation is used the improved resolution allows the small oxidation levels to be distinguished from the exponential tail. The fitting of this region, and the peak assignments will be discussed below. Some authors have suggested that there is a significant shift between the C1s from graphite (sometimes referred to as aromatic carbon) and the C1s from aliphatic carbon. In fact, there is little evidence that such a shift actually exists. For example, if one examines the C1s region of polystyrene (–CH CHPh–)n (where Ph represents a benzene ring) there #is no significant shift between the CH carbon (aliphatic carbon) and the carbon atoms# in the 3
Carbons and Graphites: Surface Properties of (a)
(b) Cls–PAN (M40) fiber
insulating hydrocarbon on conducting graphite) are in close proximity to one another.
Intensity (arb. units)
after 10 min etch
before etching
Intensity (arb. units)
achromatic
monochromatic FWHM (excluding tail) = 0.35 eV
290 285 Binding energy (eV) 293 285 277 Binding energy (eV)
Intensity (arb. units)
Figure 2 (a) Core XPS spectra of the C1s of a carbon fiber before and after exposure to argon ions. (b) Core XPS spectra of the C1s region of an untreated carbon fiber examined with and without monochromatic x radiation.
(a)
(b)
(c)
(d)
25
15
5 25 15 Binding energy (eV)
5
Figure 3 Valence band XPS spectra of four different types of carbon obtained with monochromatic Al Kα x radiation: (a) diamond; (b) crystalline graphite; (c) microcrystalline graphite; (d) glassy carbon (after McFeely et al. 1974).
benzene ring (aromatic carbon). Curve fitting of high-resolution data obtained with monochromatic radiation suggests a possible difference of only 0.2 eV (Beamson and Briggs 1992). Apparent shifts may occur, but these can easily arise from differential sample charging (e.g., see Dickinson et al. 1973) when carbon species of differing electrical conductivity (e.g., 4
(b) Valence band spectra. The valence band region of carbons shows some marked differences for different types of carbon. In the important early work of McFeely et al. (1974) an extensive study of carbon valence bands was conducted, and Fig. 3 shows the valence band region of graphite, diamond, and a glassy carbon from this work. This work pointed out that no difference would be expected in the valence band of a crystalline or a microcrystalline (polycrystalline) sample because the gross features of the density of states depend upon relatively shortrange order in the crystal. The principal difference in the valence band region appears between graphite and diamond. This difference can be understood when the spectra are compared with spectra generated by taking the density of states for different atomic orbitals from band structure calculations, and adjusting this by the photoelectron cross sections, and convoluting the result with the photon lineshape. Figure 4 compares these spectra for diamond and graphite. The calculations can be seen to reproduce the principal features of the experimental spectra, the vertical lines connecting the principal features in experimental spectra with the calculated spectra. The feature around 15 eV can be seen to be much more pronounced in the diamond spectra, and this difference is clearly predicted by the band structure calculations. Figure 10 shows the possible models that might be used for calculating carbon spectra. The reason for choosing coronene is that it is sufficiently large that it partially represents the graphite lattice but is sufficiently small that it can be functionalized to represent oxidized carbon (to be discussed below). The experimental valence band of benzene, well predicted by calculations (Riga et al. 1977, Sherwood 1992) shows sharp characteristic features. The calculated spectrum of coronene (Fig. 9(d)) shows a spectrum that is quite close to that of the graphite valence band, suggesting that the use of substituted coronenes is appropriate for predicting the valence band spectra of oxidized carbons.
2.2 Spectra of Oxidized Carbons XPS is particularly well suited to probe the surface chemistry of oxidized carbons. Oxidized carbons are important in many applications, including those where the surface of a carbon fiber is oxidized to enable better interaction between the fiber and the matrix into which the fiber is placed to make a composite material. The improved interaction arises because of improved
Carbons and Graphites: Surface Properties of
(a)
Calculated from band structure
20 10 0 Binding energy (eV)
Experimental Intensity (arb. units)
Intensity (arb. units)
Experimental
(b)
Calculated from band structure 30 20 10 0 Binding energy (eV)
Intensity (arb. units)
Figure 4 (a) Valence band XPS spectrum of graphite obtained with monochromatic Al Kα x radiation and the spectrum calculated using a band structure calculation for a twodimensional graphite lattice (after Viswanathan et al. 1997). (b) Valence band XPS spectrum of diamond obtained with monochromatic Al Kα x radiation and the spectrum calculated using a band structure calculation for a diamond lattice (the diamond experimental data is from McFeely et al. (1974), and had a nonlinear background removed; the calculation was carried out in a similar manner to the graphite calculation in Fig. 9 (P. M. A. Sherwood, unpublished results)).
290
0.5 V
2.0 V
1.0 V
2.5 V
1.5 V
3.0 V
285 290 Binding energy (eV)
285
Figure 5 Core XPS spectra of the C1s region of a polyacrylonitrilebased carbon fiber after potentiostatic oxidation to various potentials (shown versus the saturated calomel electrode) for 20 min in 2.7 mol dmV$ nitric acid solution. The spectra are fitted to six features at positions corresponding to the features in Table 1. A non-linear background is included in the fit (after Sherwood 1996a, 1996b).
wetting of the fiber surface by the matrix, and physical and chemical interaction between the oxidized fiber and the matrix. (For a more detailed discussion see, for example, Wang et al. (1998) and Sherwood (1996, 1998).) Core and valence band XPS provides valuable and complementary information that allows the functional groups on the oxidized surface to be identified, as well as an overall idea of the level of oxidation. Quantification in these systems is not possible except in the most well-defined cases. This is because quantification models in XPS require a number of assumptions whose use in carbon systems is often questionable. For example, quantification models are based upon assumptions such as a uniform homogeneous-composition flat layer of oxide on top of an elemental substrate. Except in the highest-quality graphite, carbons are rarely flat, and oxidation is rarely uniform. Variable take-off angle experiments, effective in flat systems, are of questionable value in many carbon systems such as fibers. The area ratio of different core peaks, adjusted for cross-section, electron escape depth, and the instrument transmission function can be a valuable measure for comparing the variation in the level of oxidation in a range of different samples. The C1s region shows a substantial shift for carbon\oxygen species, typical binding energies for such species being shown in Table 1. In the case of oxidized carbons it is usually necessary to use curvefitting methods to determine the positions and intensities of the overlapping peaks involved. Figure 5 shows a typical example of such a fitting approach for carbon fibers subjected to different levels of oxidation. It is important to remember that curve fitting approaches never give a unique solution. This means that it is essential to ensure that the fit is chemically reasonable, and to see if the same fitting approach can be applied to a system where the chemical composition changes in a systematic manner. In the case illustrated in Fig. 5 the level of oxidation can be seen to be increasing as a result of increasingly positive potentials to which the fibers are subjected. The fits in Fig. 5 were carried out so that the positions and widths of the fitted peak were consistent through the series of spectra. A further discussion of curve fitting in such systems can be found in a paper by Sherwood (1996b). The use of achromatic x radiation in carbon systems presents the problem that the oxidized carbon can be lost by surface decomposition in the spectrometer. This decomposition is largely caused by the heat from the target of the x-ray gun. This decomposition can be seen by substantial changes in the C1s region with increasing recording time, which are illustrated in Fig. 6 (Viswanathan et al. 1997). When monochromatic x radiation is used, where the x-ray gun is removed to a substantial distance from the sample, and high-energy Bremsstrahlung radiation is eliminated, no significant decomposition is seen. Comparison of the changes in the C1s and valence band region for different x-ray 5
Carbons and Graphites: Surface Properties of Monochromatic
(a)
Achromatic
(b)
0.2/1.5 h
Intensity (arb. units)
20 min
2h
Intensity (arb. units)
1.3/7 h
3/28 h
20 h
290
285
290
24/48 h
285
Binding energy (eV)
Figure 6 Core XPS spectra of the C1s region of a polyacrylonitrile based carbon fiber after galvanostatic oxidation in 1 mol dmV$ nitric acid solution at 0.5 A for 40 s studied using achromatic and monochromatic x-radiation for different exposure times (after Viswanathan et al. 1997).
Table 1 C1s features for different carbon\oxygen functionalities Peak No.
Functionality
Binding energy (eV )
Shift (eV )
1
Graphitic carbon
284.6
0
2
B carbon
285.3
0.7
3
C OH
286.1
1.5
4
Bridged structure
286.6–286.9
2.0–2.3
5
C O
287.6–288.2
3.0–3.6
6
COOH, COOR
288.8
4.2
7
CO#V $
289.1
4.5
8
π–π*
290.7
6.1
9
Plasmon
291.5
6.9
6
290
284 33 Binding energy (eV)
3
Figure 7 Core XPS spectra of (a) the C1s region and (b) the valence band XPS spectra of pitch-based carbon fibers after potentiostatic oxidation at 3.0 V (SCE) in 1 mol dmV$ nitric acid solution. Times of x-ray exposure are indicated on the right of the figure, the first number referring to the C1s region and the second number to the valence band region (after Xie and Sherwood 1990a, 1990b).
exposure times using achromatic radiation indicate that the C1s region shows the largest change in spectral appearance due to decomposition (Xie and Sherwood 1990b). The higher kinetic energy in the valence band region (about 1230 eV with Mg Kα x rays) compared to the C1s region (about 960 eV with Mg Kα x rays) would suggest that the latter region is more surface sensitive. However, initial inspection of the spectra can be misleading because the valence band region is composed of regions of significant oxygen and carbon intensity over a wide energy range. Thus, it is tempting to focus on the principal oxygen feature, the O2s region. The ratio of the O2s area (the peak around 25 eV in Fig. 7) to the total area in the valence band
Carbons and Graphites: Surface Properties of
12.34 8.67 7.04
10.58
6.74
8.92
4.00
8.85 5.85
2.59
8.77 5.45
2.27
Binding energy (eV) Figure 8 The calculated XPS valence band spectra of various substituted coronenes using multiple scattered wave Xα calculations (after Sherwood 1996a, 1996b).
region falls from about 0.32 to about 0.27 over the exposure times in Fig. 7 compared with the fall in the oxygen\carbon atomic ratio from the O1s\C1s peak areas from about 0.6 to about 0.4 for the same time. These differences will be discussed in more detail below, when it will be seen that the changes in the valence band region do correspond to significant decomposition, comparable to that seen in the C1s region. The valence band region provides complementary information to that found in the core region. Calculations for different functional groups on carbon using coronene with different types of functionality (Xie and Sherwood 1991a) provide a good prediction of what can be expected. Figure 8 shows that differences would be expected in the number of peaks in the principally O2s region around 25 eV (one for all the calculated functionalities except carboxyl which would be
expected to show two peaks), and in the region below 16 eV which shows a characteristic ‘‘fingerprint.’’ The position of the principally O2s region will be largely determined by the extent of interaction between O2s and C2s orbitals, which is different from the largely electrostatic effects that give rise to shifts in the core XPS regions. Thus, it is not surprising to find that the valence band region would behave differently than the core regions. This is predicted by the calculations, so while the C1s core region would be expected to give the same shift for hydroxide and epoxide functionalities, the valence band regions would be expected to be different, a prediction which is found to occur in spectra of oxidized carbon, where both of these species would be expected to be present (Xie et al. 1992). It is helpful to separate out the oxygen and carbon contributions to the valence band region. This has been done (P. M. A. Sherwood, unpublished results) in 7
Carbons and Graphites: Surface Properties of
(c)
(b)
(d)
Intensity (arb. units)
(a)
30
20
10
0 30 Binding energy (eV)
20
10
0
Figure 9 The calculated XPS valence band spectrum of a coronene and substituted coronene. The spectra in (a), (b) and (c) are for the ‘‘bridged’’ structure (Fig. 10), and the spectrum in (d) is for coronene. The substituted coronene spectrum for C O H #% (c) "# ' is shown in (a). Spectrum (b) shows the part of the spectrum in (a) corresponding to O2s and O2p intensity. Spectrum shows the part of the spectrum in (a) corresponding to C2s and C2p intensity. Spectra (b) and (c) are not adjusted for scale. The total area of spectrum (b) is 73% of the area of the total spectrum, and spectrum (c) is 27% of the area of the total spectrum.
Fig. 9 for the substituted coronene with the bridged structure (Fig. 10). The component peaks in the figure correspond to the position of all the energy levels in the valence band region with an area corresponding to the atomic orbital contribution adjusted by the photoelectron cross-section and a FWHM set at 2.58 eV. In fact, it is known that the widths of the component levels should vary within a spectrum due to effects such as vibrational broadening, but these are not easily predicted so all the peaks are set at the same width for predictive purposes. It is clear that the oxygen contribution (Fig. 9(b)) makes the principal contribution to the spectrum (73% of the intensity for C O H ). #%The"# heights ' of the spectra in Fig. 9 have all been adjusted so that the peak maxima are about the same for display purposes; if drawn to scale, Figs. 9(c) and 8
(d) would be of much lower intensity. It is also clear that the interaction between O2s and C2s electrons in the bridged structure leads to a considerable change in the distribution of carbon intensity when Fig. 9(c) (carbon contribution) is compared with the graphite ‘‘model’’ system coronene (C H ) in Fig. 9(d). Thus, the O2s region is referred to#% as"#being ‘‘principally’’ O2s above, because Fig. 9(c) indicates that significant C2s intensity moves into this region as a result of O2s–C2s interactions. There are also significant differences in the largely C2p region below 16 eV for Figs. 9(c) and (d), indicating that O2p–C2p interactions significantly impact spectral features. Hydrogen has such a small cross-section that it makes negligible contributions to all these spectra. Another important point that arises is that the principally O2s region around 25 eV binding
Carbons and Graphites: Surface Properties of
Benzene
Coronene Substituted coronene (bridged structure)
Figure 10 Models for calculating carbon spectra.
energy accounts for only 44% of the total oxygen intensity in the spectrum even though it is the most obvious feature in the spectrum. One can use the spectra in Fig. 9 to generate spectra for different carbon\oxygen atomic ratios. For example, Fig. 11 shows the valence band region for an oxidized carbon fiber obtained with monochromatic x radiation (which eliminates sample decomposition), for which the carbon\oxygen atomic ratio obtained from the C1s\O1s area ratios adjusted to give an atomic ratio as described above is 3. The calculated spectrum was obtained by first adjusting Fig. 9(b) to increase the width of peaks in the O2s region (the first seven energy levels of the 58 valence energy levels in the calculation) from a FWHM of 2.58 to 5.94 eV while keeping the peak areas the same (so keeping the contributions from each component energy level the same) to better reflect the observed widths of this region. Then the modified Fig. 9(b) was added to Fig. 9(c) and two times Fig. 9(d). This addition was performed in this way to ensure that one recognized that the C1s intensity in Fig. 9(c) results from carbon bonded to oxygen in the highly oxidized coronene shown in Fig. 10, and that unoxidized carbon is better represented by the coronene spectrum in Fig. 9(d). The agreement between the calculated and experimental spectra in Fig. 11 can be seen to be reasonable. The valence band calculation helps one reconcile the information provided in the core and the valence band regions. Figure 11 shows that when the core and valence band regions are recorded with monochromatic x radiation (when no decomposition is expected), then the carbon\oxygen atomic ratio is consistent for the core and valence band regions. This calculation helps us to reconcile the apparent observed
differences in the oxygen\carbon atomic ratio in the core region and the change in the O2s percentage in the valence band regions discussed above for the data in Fig. 7. When the ratio of the O2s area (the peak around 25 eV in Fig. 7) to the total area in the valence band region is adjusted to reflect the fact that some 66% of the oxygen intensity lies outside this region, and other helpful data from the calculation are included, an oxygen\carbon ratio can be obtained for the valence band region. The calculation indicates that for an oxygen\carbon ratio of 1, 81.5% of the valence band region will correspond to oxygen intensity, and 17% of the carbon intensity lies in the O2s region. Of course, one could obtain the first value from the photoelectric cross-section ratios (using cross-sections from Scofield (1976)), but such a calculation would require the assumption of a hybridization state for carbon and oxygen which would be necessarily more approximate than using the data from the calculation. An approximate measure of the oxygen\carbon ratio is obtained by taking the fraction of the area of the valence band region represented by the O2s region (let us call this x) and adjusting it by 1\0.44 to give the fraction of the valence band due to oxygen (2.27x). If we note that for an oxygen\carbon ratio of 1, about 3% of the O2s region is carbon intensity, then one can adjust this fraction to 2.35x for an average oxygen\ carbon ratio of 0.5 over this region. The difference between this value and the total area (1V2.35x) will represent the area due to carbon, which is then adjusted by (1\0.185) to reflect that fact that cross-section differences mean that when there is a oxygen\carbon ratio of 1, 81.5% of the valence band region will correspond to oxygen intensity. When these adjustments are performed, the oxygen\carbon 9
Carbons and Graphites: Surface Properties of
Intensity (arb. units)
Experiment
the substituted coronene of Fig. 10 linearly on the surface, then the thickness of oxidized carbon would be 1.28 AH (the C O bond length) and the oxygen\ carbon ratio from the valence band region should be 0.88 times that from the C1s region if the inelastic mean free path is 20 AH in the C1s region and 22.6 AH in the valence band region. Using these figures, a range of 0.62–0.37 from the C1s region should correspond to a range of 0.55–0.33 in the valence band region, very close to the observed figures listed above.
3. Concluding Comments
Calculated for C/O ratio = 3
Carbon presents some special challenges in the determination of its surface chemistry and topography, but the understanding of these properties is essential for the evaluation and understanding of many important properties of carbons. Core and valence band XPS techniques are probably the most valuable methods for the nondestructive evaluation of carbon surface chemistry, and AFM, STM, and electron microscopy are effective for the determination of surface topography.
Acknowledgments This material was based upon work supported by the National Science Foundation (under Grant No. CHE9421068), NASA, and the State of Kansas. 31
23
15
7
Binding energy (eV)
Figure 11 Comparison of the calculated XPS valence band spectrum of polyacrylonitrile-based carbon fibers after galvanostatic oxidation in 1 mol dmV$ nitric acid solution at 0.25 A for 40 s (experimental data from Viswanathan et al. 1997). The calculated spectrum was performed by combining the data in such a way that the carbon\oxygen ratio was 3 (in agreement with the experimental evaluation based on the C1s to O1s intensity ratio). The way in which the calculated spectrum was generated is explained in the text.
ratio in the valence band region of Fig. 7 ranges from 0.56 to 0.32. The change in the oxygen\carbon ratio calculated in this way is thus more striking than the change in the percentage of the valence band in the O2s region (0.32–0.27), indicating how a visual inspection of the spectral features can be misleading. This calculated atomic ratio is in line with the oxygen\carbon ratio range from the C1s and O1s region, namely 0.62–0.37. The greater surface sensitivity of the C1s region would lead one to expect a greater oxygen\carbon ratio using C1s rather than valence band data. For example, if one had a flat carbon surface that consisted of the functionality of 10
Bibliography Beamson G, Briggs D 1992 High Resolution XPS of Organic Polymers—The Scienta ESCA300 Database. Wiley, Chichester, UK Dickinson T, Povey A F, Sherwood P M A 1973 Differential sample charging in ESCA. J. Electron Spectrosc. Relat. Phenom. 2, 441–7 Doniach S, SB unjic! 1970 Many-electron singularity in x-ray photoemission and x-ray line spectra from metals. J. Phys. C 3, 285–91 McFeely F R, Kowalczyk S P, Ley L, Cavell R G, Pollak R A, Shirley D A 1974 X-ray photoemission studies of diamond, graphite, and glassy carbon valence bands. Phys. Re. B 12, 5268–78 Riga J, Pireaux J J, Verbist J J 1977 An ESCA study of solid benzene. Mol. Phys. 34, 131–8 Rooke M A, Sherwood P M A 1994 x-ray photoelectron spectroscopic study of the effect of ion etching of silicon carbide on a carbon fiber. Carbon 33, 375–80 Scofield J H 1976 Hartree–Slater subshell photoionization crosssections at 1254 and 1487 eV. J. Electron Spectrosc. Relat. Phenom. 8, 129–37 Sherwood P M A 1992 Carbon, functionalized carbon, and hydrocarbons studied by valence band photoemission. J. Vac. Sci. Technol. A 10, 2783–7 Sherwood P M A 1996a Surface analysis of carbon and carbon fibers for composites. J. Electron Spectrosc. 81, 319–42 Sherwood P M A 1996b Curve fitting in surface analysis and the
Carbons and Graphites: Surface Properties of effect of background inclusion in the fitting process. J. Vac. Sci. Technol. A 14, 1424–32 Sherwood P M A 1998 Composites. In: Rivie' re J C, Myhra S (eds.) Handbook of Surface and Interface Analysis. Dekker, New York, pp. 605–41 Thoms B D, Russel J N Jr., Pehrsson P E, Butler J E 1994 Adsorption and abstraction of hydrogen on polycrystalline diamond. J. Chem. Phys. 100, 8425–31 Viswanathan H, Rooke M A, Sherwood P M A 1997 X-ray photoelectron spectroscopic studies of carbon fiber surfaces. 21 Comparison of carbon fibers electrochemically oxidized in acid using achromic and monochromatic XPS. Surface Interface Anal. 25, 409–17 Wang T K, Donnet J-B, Peng J C M, Rebouillat S 1998 In: Donnet J-B, Wang T K, Peng J C M, Rebouillat S (eds.) Carbon Fibers, 3rd edn. Dekker, New York, pp. 231–309 Weaver J H 1992 Fullerenes and fullerides: photoemission and scanning tunneling microscopy studies. Acc. Chem. Res. 25, 143–9 Wei J, Yates J T Jr. 1995 Diamond surface chemistry I—a review. Diamond surface chemistry II—selected bibliography. Crit. Re. Surface Chem. 5, 1–248 Xie Y, Sherwood P M A 1990a x-ray photoelectron spectroscopic studies of carbon fiber surfaces. Part XII: the effect of
microwave plasma treatment on pitch-based carbon fiber surfaces. Appl. Spectrosc. 44, 797–803 Xie Y, Sherwood P M A 1990b X-ray photoelectron spectroscopic studies of carbon fiber surfaces. Part XIV: electrochemical treatment of pitch-based fibers and the surface and bulk structure changes monitored by XPS, XRD, and SEM. Appl. Spectrosc. 44, 1621–8 Xie Y, Sherwood P M A 1991a X-ray photoelectron spectroscopic studies of carbon fiber surfaces. 13. Valence-band studies of oxidized fibers interpreted by xα calculations. Chem. Mater. 3, 164–8 Xie Y, Sherwood P M A 1991b X-ray photoelectron spectroscopic studies of carbon fiber surfaces. Part XV: electrochemical treatment on pitch-based fibers by potentiostatic and galvanostatic methods. Appl. Spectrosc. 45, 1158–65 Xie Y, Wang T, Franklin O, Sherwood P M A 1992 X-ray photoelectron spectroscopic studies of carbon fiber surfaces. Part XVI: core-level and valence-band studies of pitch-based fibers electrochemically treated in ammonium carbonate solution. Appl. Spectrosc. 46, 645–51 Yin M, Koutsky J A, Barr T L, Rodriguez N M, Baker R T K, Klebanov L 1993 Characterization of carbon microfibers as reinforcement for epoxy resins. Chem. Mater. 5, 1024–31
P. M. A. Sherwood
Copyright ' 2001 Elsevier Science Ltd. All rights reserved. No part of this publication may be reproduced, stored in any retrieval system or transmitted in any form or by any means : electronic, electrostatic, magnetic tape, mechanical, photocopying, recording or otherwise, without permission in writing from the publishers. Encyclopedia of Materials : Science and Technology ISBN: 0-08-0431526 pp. 985–995 11