XPS studies of graphite electrode materials for lithium ion batteries

Applied Surface Science 167 Ž2000. 99–106 www.elsevier.nlrlocaterapsusc

XPS studies of graphite electrode materials for lithium ion batteries R.I.R. Blyth a,) , H. Buqa b, F.P. Netzer a , M.G. Ramsey a , J.O. Besenhard b, P. Golob c , M. Winter b a

b

Institut fur 5, A-8010 Graz, Austria ¨ Experimentalphysik, Karl-Franzens UniÕersitat ¨ Graz, UniÕersitatsplatz ¨ Institut fur ¨ Chemische Technologie Anorganischer Stoffe, Technische UniÕersitat ¨ Graz, Stremayrgasse 16, A-8010 Graz, Austria c Forschungsinstitut fur ¨ Elektronenenmikroscopie und Feinstrukturforschung, Technische UniÕersitat ¨ Graz, Steyrergasse 17, A-8010 Graz, Austria Received 5 April 2000; accepted 1 July 2000

Abstract Surface pre-treatment of graphitic electrode materials for lithium ion cells has recently been shown to significantly reduce the irreversible consumption of material and charge due to the formation of the so-called solid electrolyte interphase ŽSEI. during battery charging. In this paper, we compare graphite powders and carbon fibres as model materials for X-ray photoemission spectroscopy ŽXPS. studies of the effects of surface pre-treatments. For carbon fibres, the surface carbon percentage was found to vary from 70–95% depending on the surface treatment, with corresponding changes in the relative proportion of graphitic compared to C`O bonds, as determined from C 1s curve fits. In contrast, results from the graphite powders show very little change in surface chemical composition and an essentially constant C 1s lineshape dominated by graphitic carbon. SEM data show the carbon fibre cross-section to be composed of a radial array of layered graphite, leaving a surface consisting largely of prismatic planes, while the graphite powder consists of graphite platelets with the surface area predominantly of basal planes. We conclude that the chemical modification occurs at the prismatic planes, and that the powders are unsuitable as models for XPS studies of electrode surface modification, while the fibres are very well suited. q 2000 Elsevier Science B.V. All rights reserved. PACS: 81.05.T; 79.60.Fr Keywords: X-ray photoemission spectroscopy; Graphite powder; Carbon fibre; Lithium ion batteries; Carbon electrodes

1. Introduction Lithium ion cells have a large and increasing share of the rechargeable battery market. Since these

) Corresponding author. Tel.: q43-316-380-5219; fax: q43316-389-9816. E-mail address: [email protected] ŽR.I.R. Blyth..

cells typically operate at voltages beyond the thermodynamic stability range of the organic electrolytes, electrolyte decomposition occurs. However, the resulting reaction products can form films at the electrodes, which, in an ideal case, produce an electrically insulating layer, stopping further decomposition, while still permitting the transport of lithium ions. At the anode, such a film is referred to as a solid electrolyte interphase ŽSEI. w1,2x. SEI forma-

0169-4332r00r$ - see front matter q 2000 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 9 - 4 3 3 2 Ž 0 0 . 0 0 5 2 5 - 0

100

R.I.R. Blyth et al.r Applied Surface Science 167 (2000) 99–106

tion inevitably results in the partial consumption of both electrolyte and lithium, leading to a loss in charge known as the irreversible capacity, Cirr . This reduces the specific energy of the cell, and increases the required material expenses for cell construction, and it is therefore advantageous to be able to reduce Cirr while still forming an effective SEI. The performance of lithium ion cells depends strongly on the type of electrode materials used, with graphitic carbons emerging as the material of choice for the anode through a combination of electrochemical, economic and environmental reasons w1x. Graphite possesses two kinds of surfaces: the basal plane and the prismatic Žedge. surfaces. The basal plane of graphite consists largely of carbon atoms. In contrast, the considerably more open prismatic planes include various, mostly oxygen- and hydrogen-containing surface groups. The surface chemistry and morphology of the prismatic surfaces of graphite play a major role in chemical and electrochemical reactivity, interaction with SEI products, etc. w1–3x. Moreover, the transport of lithium ions during battery cycling largely takes place via the prismatic planes rather than the basal planes. In conclusion, the amount, morphology and chemical composition of prismatic surfaces have a significant impact on the anode performance. One route to the reduction of Cirr is to chemically treat the surface of the graphite prior to electrode construction. For example, we have recently shown that pre-treatment of graphite electrode materials with gaseous O 2 and CO 2 at elevated temperatures produces a substantial reduction in Cirr w3x. Whereas ideal basal planes are rather unreactive, the surface chemistry of the prismatic planes may be changed significantly during the gaseous treatment at elevated temperatures, and anode performance in turn may be changed as well w3x. Therefore, the chemical state of the exposed prismatic planes is of considerable interest. In principle, X-ray photoemission spectroscopy ŽXPS. measurements can be used to provide direct information on the surface chemical state of the exposed prismatic planes, but care has to be taken in the choice of graphitic material to be studied. The material chosen must permit XPS analysis of at least a significant proportion of prismatic surface area, and it must also be suitable for the construction of Žat minimum, prototype. battery electrodes with

which Cirr can be determined. There are several different types of graphitic material available. Natural, single-crystal graphite has been the subject of numerous electron spectroscopic studies, which include XPS w4,5x, angle resolved photoemission w6x and inverse photoemission w7x. It cleaves readily in UHV, but of course, along the basal planes which are not of interest for our purposes, while cleavage along the prismatic planes is a great deal more problematic. SEI formation on highly oriented pyrolitic graphite ŽHOPG. has been recently investigated using XPS, with significant differences found between the SEI formed on the basal and prismatic planes w8x. Two further graphitic materials, both of which are particularly suited to electrochemical investigations, and the construction of prototype electrodes, are graphite powders and carbon fibres. Graphite powders, with particle sizes around the 10-mm range, were used in the CO 2 and O 2 studies reported earlier w3x. While there is an extensive literature on the surface properties of carbon fibres w9x, XPS studies on graphitic powders are much less common. Such studies as do exist concern the so-called carbon blacks w10x, with particle sizes in the tens of nm range, and a great deal of disorder. In this paper, we compare the suitability of graphite powders and carbon fibres for XPS studies of surface pre-treated electrodes. Four different treatment procedures were followed for both powder and fibre samples treatment with CO 2 and O 2 , and two different proprietary treatments, which we shall call P1 and P2 . Although the details of treatments P1 and P2 cannot at present be given, both were followed identically for powder and fibre samples. As will be shown, treatments P1 and P2 produce more dramatic surface chemical modification than either CO 2 or O 2 . This allows a more extensive comparison of the suitability of these graphitic materials for XPS studies than if the proprietary treatments were excluded from this study.

2. Experimental The XPS measurements were performed with a PHI 5400 electron spectrometer, using unmonochromated Mg K a radiation Ž hn s 1253.6 eV., giving

R.I.R. Blyth et al.r Applied Surface Science 167 (2000) 99–106

an overall energy resolution of ; 1 eV for survey scans and 0.8 eV for the C 1s spectra. The electron emission angle was 308. Ninety-five percent of the observed elastic photoelectron signal comes from a layer whose thickness, d, is given by w11x d s 3 lsin a

Ž 1.

where a is the emission angle and l is the attenuation length of electrons in a solid, which in turn depends on the kinetic energy of those electrons. For the case of the C 1s level excited using Mg K a radiation, E K ; 960 eV, l is about 1.2 nm w11x, which gives a value for d of 1.8 nm at an emission angle of 308. The vacuum system, which was equipped with a load lock such that the sample could be introduced into vacuum without being baked, had a base pressure of 1 = 10y1 0 mbar. The carbon fibre samples were clamped to a gold-plated sample holder, while the graphite powder samples were mounted using double-sided conducting adhesive tape. The SEM data were taken using a LEO DSM 982 Gemini scanning microscope using an electron energy of 5 keV. The graphite powders used were Timrex w special graphites SFG 44, as used in the Cirr studies w3x and SFG 6 ŽTimcal, Switzerland.. These powders differ only in the particle size distribution, with 50% of the particles smaller than 22 and 3.2 mm, with specific surface areas of 4.2 and 15.2 m2rg, for SFG 44 and SFG 6, respectively Žmanufacturer’s figures.. The carbon fibres used were Thornel w P100S pitch based carbon fibres ŽAmoco.. Commercially available carbon fibres for application in carbon fibrerpolymer composites, such as the P100S fibres, are industrially surface treated in order to provide good adhesion of the fibre with the polymer. The detailed methods of surface treatment are proprietary and not disclosed in the literature. However, basically, the industrial surface treatment involves an oxidation step and a subsequent polymer treatment step w12x. To remove the polymer from the surface, the P100S fibres were first extracted in acetone in a Soxhlet apparatus for 48 h, then dried in low vacuum at 1508C for 4 h and finally heat-treated in argon as described below. The CO 2 and O 2 treated samples were prepared in a two-step procedure as follows. The carbon was placed in quartz glass tube and heated in a tube

101

furnace under an argon atmosphere from room temperature to 10008C in 208Crmin steps. The sample was held at 10008C for 1 h in order to remove the pristine surface-groups and impurities from the carbon surface. After this AcleaningB step, the carbon was cooled to ; 5008C and treated with CO 2 or O 2 for 15 min and 2 h, respectively. After treatment, the tube was cooled to room temperature under argon atmosphere, and the samples were stored under dry conditions. For comparison, samples of both fibre and powder were prepared by cleaning in argon at 10008C for 1 h without subsequent gas treatment.

Fig. 1. SEM micrographs showing the cross-section of a P100S fibre and a particle of SFG 44 graphite powder.

R.I.R. Blyth et al.r Applied Surface Science 167 (2000) 99–106

102

3. Results and discussion SEM micrographs of untreated materials are shown in Fig. 1. The SFG 44 powder consists of crystalline graphite platelets, which have a relatively small proportion of exposed prismatic faces compared to basal planes. The P100S fibre is rather different, as its cross-section in Fig. 1 shows. It consists of a radial arrangement of graphite planes, such that the exterior of the fibre, i.e. the surface as seen by XPS, is dominated by the prismatic planes containing surface groups. The surface elemental composition was determined from the relative areas of the C 1s and O 1s peaks, no other elements being detected after the subtraction of the inelastic background, and taking into account their photoemission cross-sections for Mg K a radiation w13x. The results are shown in Table 1. Note that, since the samples had been exposed to atmosphere, the contribution of contaminants to the C and O signal cannot be ruled out. The carbon percentage for the treated SFG 44 powders remains essentially constant at 96 " 1.5% compared to 99% for the powder after cleaning in argon alone. An exception is the P2 treated powder, which has a carbon percentage of 89.9%. In contrast, the carbon percentage for the P100S fibres varies widely, from nearly 95% for the CO 2 treated fibre to 70% for the P2 treated fibre. The only surface treatment to produce a significant change in the surface chemical composition of the SFG 44 powder is that which produces the lowest carbon and highest oxygen percentage for the P100S fibre, i.e. proprietary treatment P2 . The argon seems to have little effect on the

Table 1 Surface elemental composition of the treated P100S fibres and SFG 44 graphite powder determined from XPS Treatment

Untreated Argon only O2 CO 2 P1 P2

P100S

SFG 44

C%

O%

C%

O%

85.0 86.6 87.2 94.7 71.4 70.5

15.0 13.4 12.8 5.3 28.6 29.5

97.4 99.01 95.6 97.1 95.8 89.9

2.6 0.99 4.4 2.9 4.2 10.1

Fig. 2. Carbon 1s XPS spectra normalised to maximum intensity of untreated and treated P100S carbon fibres.

elemental surface composition of either the carbon fibres or powders, with the differences seen being too small to be regarded as significant. The corresponding carbon 1s spectra for the P100S fibres and SFG 44 powder are shown in Figs. 2 and 3, respectively. The spectra from the P100S fibres show a wide variation in the C 1s lineshape, with the proportion of graphitic carbon at 284.4 eV w4,5x compared to the different C`O bonds, at higher binding energies, being far from constant. The lineshape of the untreated sample is rather different to

R.I.R. Blyth et al.r Applied Surface Science 167 (2000) 99–106

Fig. 3. Carbon 1s XPS spectra normalised to maximum intensity of untreated and treated SFG 44 graphite powders.

that of the argon cleaned fibre, being much broader, although the elemental analysis discussed earlier showed very little difference. In contrast to the fibre data, Fig. 3 shows that there is a very little difference between the spectra from the differently treated SFG 44 powders, being in all cases dominated by the graphitic peak. The SFG 44 spectra of Fig. 3 show some apparent peak binding energy shifts, but these are due to slight, irreproducible charging of the sample. To further quantify the relative proportions of graphitic and non-graphitic carbon present, curve

103

fitting was employed. The peak lineshapes used were Voigt functions, i.e. Gaussian–Lorentzian convolutions, with an integral background subtracted prior to fitting. The C 1s binding energy for graphite is well known to be 284.4 eV, as determined from XPS w4x and synchrotron radiation photoemission w5x measurements of single crystal graphite. The lineshape of the graphitic peak is generally regarded as being asymmetric w14x — although the high resolution data w5x show a nearly symmetric peak — with the extent of the asymmetry determined by the average area of conjugation w15x. In principle, since graphite has a finite density of states at the Fermi level, DOSŽ EF ., it is possible to excite the electron-hole pairs responsible for asymmetry in photoemission peaks. However, DOSŽ E F . is extremely low in graphite w6,7x and so the asymmetry would not be expected to be particularly significant. The C 1s binding energy for hydrocarbons, i.e. C bonded to C and H, is 285.0 eV w16x, which would of course place it directly to the high binding energy side of the graphitic peak. This means that any C`H bonds, or adventitious hydrocarbons, will tend to induce an apparent asymmetry in the graphitic C 1s peak towards higher binding energy. For simplicity in our curve fits, we have used a symmetric graphite peak and included a 285.0-eV contribution, which will take account of both any asymmetry and the presence of hydrocarbons. Note, however, that should any asymmetry be present, the fits will overestimate the hydrocarbon contribution. The large number of XPS studies of carbon fibres w9x ensures that there is a considerable literature regarding the binding energies of C 1s components, Table 2 Peak assignments for the C 1s components used in curve fitting Notation

Binding energy ŽeV.

Assignment

Graphite Hydrocarbon Oxide I

284.4 285.0 286.1–286.3

Oxide II Oxide III

287.6–287.7 288.6–289.1

Oxide IV

290.5–290.8

Graphite ŽC ` C. Hydrocarbons ŽC ` H. Alcohol ŽC ` OH.r ether ŽC ` O ` C. Carbonyl ŽC5O. Carboxyl ŽCOOH.r ester ŽCOOR. Carbonate ŽCO 3 .r satellite Ž P – P ) .

R.I.R. Blyth et al.r Applied Surface Science 167 (2000) 99–106

104

Fig. 4. Example of the C 1s spectrum of an SFG 44 sample with the results of curve fitting, showing the different chemical states and the p–p ) shake-up satellite.

due to various possible C`O bonds on the surfaces of graphitic materials. The generally accepted assignments are due to Xie and Sherwood w17x, and are shown in Table 2. These assignments are consistent with those determined recently from a study of a wide variety of polymers w16x, model materials containing known and well-defined C`O bonds, except that they are ; 0.3 eV higher in energy, presumably due to the different screening environment in Žconducting. graphite compared to Žlargely insulating. polymers. For the SFG 44 powders, a single peak Žat 286.8 eV. was used for the oxide I and oxide II peaks, as they could not be resolved in the data. A typical curve fit result for SFG 44 is shown in Fig. 4, and the numerical results of curve fitting in Tables 3 and 4 for the P100S fibres and SFG 44 powders, respectively.

The graphiticrhydrocarbon percentage, taken together because of the uncertainties regarding peak asymmetry discussed earlier, for all the SFG 44 powders is 83.0 " 1.2%, i.e. essentially constant. The proportions of the different C`O bonded carbons are low in all cases below 10% and again essentially constant. The peaks in Fig. 3, in which the different C`O peaks are hardly discernible by eye, are being fitted with a lineshape containing 20 parameters Žfive peaks, each with energy, intensity, width and Gaussian–Lorentzian mix.. The parameter space in which a good fit can be obtained is therefore rather large, and as a result, the differences seen in the results of Table 4 cannot be regarded as significant. In contrast, the graphiticrhydrocarbon percentage for the P100S fibres varies wildly, from

Table 3 Curve fit results for the P100S carbon fibre data of Fig. 2

Table 4 Curve fit results for the SFG 44 artificial graphite powder data of Fig. 3

Treatment

Graphiter hydrocarbon

Oxide I

Oxide II

Oxide III

Oxide IV

Treatment

Graphiter hydrocarbon

Oxide IrII

Oxide III

Oxide IV

Untreated Argon O2 CO 2 P1 P2

55.7 60.9 78.1 65.0 18.7 4.4

29.8 23.6 7.2 28.1 39.9 59.1

6.0 6.7 6.3 2.2 18.7 15.5

4.8 5.2 6.7 1.8 6.3 5.9

3.7 3.4 4.1 2.9 16.2 14.9

Untreated Argon O2 CO 2 P1 P2

84.9 85.4 83.9 81.8 83.5 82.1

4.9 6.4 7.2 9.0 7.2 7.5

3.6 3.1 3.3 3.2 4.4 5.2

6.5 5.1 5.5 5.9 4.9 5.1

R.I.R. Blyth et al.r Applied Surface Science 167 (2000) 99–106

4% to almost 80%, with a corresponding wide distribution among the possible C`O bonds. Unsurprisingly, the treatment, which produces the lowest percentage of graphiterhydrocarbon, is that which produces the lowest carbon and highest oxygen relative concentrations, i.e. proprietary treatment P2 . The curve fit results for the untreated and argon cleaned fibres as shown in Table 3 are very similar. The major difference is due to the balance between hydrocarbon and graphitic carbon contributions. For the untreated fibre, the percentages of graphite and hydrocarbon are 20.6% and 34.8%, respectively, i.e. there is almost twice as much hydrocarbon visible as graphite, although the hydrocarbon contribution may be somewhat overestimated. The result is the relatively broad peak seen in Fig. 2, with binding energy close to the hydrocarbon value of 285.0 eV. After argon cleaning, the graphite and hydrocarbon percentages are 31.2 and 29.7, respectively, i.e. essentially equal, resulting in the narrower peak at 284.4 eV seen in Fig. 2. This shows that the principal effect of heating the P100S fibres in argon is the removal of hydrocarbons, which can be assumed to stem from the polymer coating. The results suggest that the chemical modifications are taking place at the prismatic planes, given their visibility in the P100S data, and not at the basal planes, given their absence in the SFG 44 data. In

Fig. 5. Carbon 1s XPS spectrum of untreated and P2 treated SFG 6 graphite powder.

105

order to determine whether increasing the ratio of prismatic to basal planes would improve the chances of observing changes in the C 1s spectra, we also measured XPS spectra from SFG 6 powder, which has a much smaller average particle size and higher active surface area as described earlier. This should result in a larger percentage of prismatic planes. Treatment P2 was used as it gives the most dramatic changes in chemical and elemental composition in the P100S data. The C 1s spectra from untreated and P2 treated SFG 6 powder are shown in Fig. 5, and as can be clearly seen, there are no observable changes in the C 1s lineshape.

4. Conclusions There are two main conclusions to be drawn from these data. Firstly, the surface chemical modifications induced by the various treatments are indeed taking place preferentially at the prismatic planes. This is consistent with the C 1s XPS results for Žair exposed. HOPG w8x, where the prismatic surface shows significant amounts of oxide I, while the basal plane shows largely graphitic and hydrocarbon signal, with little or no oxide evident. We note in passing that the peak assignments used in the HOPG study w8x are rather different to those in this work, with in particular, the binding energies of hydrocarbons, carbonyl and alcohol groups inconsistent with the literature consensus w17x. Secondly, the proportion of prismatic planes visible to XPS for the graphite powders is simply too low for any effects to be determined unambiguously. This remains true even when using the SFG 6 material with its smaller particle size and thus, higher proportion of prismatic surface area. We conclude that the SFG 44 Žand SFG 6. powders are not particularly suited to XPS studies relevant to SEI formation, although they are the material used in actual graphite electrodes. However, the fibres provide a good model, due to the much larger proportion of prismatic surfaces visible to XPS. In addition, the fibres can also be used for electrochemical studies of SEI formation, and a reduction in Cirr has in fact been observed w18x after the same CO 2 and O 2 treatments, which reduced Cirr for powder electrodes w3x.

106

R.I.R. Blyth et al.r Applied Surface Science 167 (2000) 99–106

Acknowledgements We would like to thank the SFB AElektroaktive StoffeB of the Austrian Science Foundation for financial support, and Jorg ¨ Gomm and Hartmut Schrottner for invaluable technical work. The ¨ graphite powders were donated by Timcal, Bodio, Switzerland.

References w1x M. Winter, J.O. Besenhard, in: J.O. Besenhard ŽEd.., Handbook of Battery Materials, Wiley-VCH, Weinheim, 1998. w2x E. Peled, D. Golodnitzky, J. Penciner, in: J.O. Besenhard ŽEd.., Handbook of Battery Materials, Wiley-VCH, Weinheim, 1999. w3x M. Winter, H. Buqa, B. Evers, T. Hodal, K.-C. Muller, C. ¨ Reisinger, M.V. Santis Alvarez, I. Schneider, G.H. Wrodnigg, F.P. Netzer, R.I.R. Blyth, M.G. Ramsey, P. Golob, F. Hofer, C. Grogger, W. Kern, R. Saf, J.O. Besenhard, ITE Batt. Lett. 1 Ž1999. 129. w4x P.M.Th.M. van Attekum, G.K. Wertheim, Phys. Rev. Lett. 43 Ž1979. 1896.

w5x F. Sette, G.K. Wertheim, Y. Ma, G. Meigs, S. Modesti, C.T. Chen, Phys. Rev. B 41 Ž1990. 9766. w6x A.R. Law, M.T. Johnson, H.P. Hughes, Phys. Rev. B 34 Ž1986. 3289. w7x I.R. Collins, P.T. Andrews, A.R. Law, Phys. Rev. B 38 Ž1989. 13348. w8x D. Bar-Tow, E. Peled, L. Burstein, J. Electrochem. Soc. 146 Ž1999. 824. w9x P.M.A. Sherwood, J. Electron Spectrosc. Relat. Phenom. 81 Ž1996. 319. w10x E. Papirer, R. Lacroix, J.B. Donnet, Carbon 34 Ž1996. 1521, and references therein. w11x P.M.A. Sherwood, in: D. Briggs, M.P. Seah ŽEds.., Practical Surface Analysis, 2nd edn., Wiley, Chichester, 1990. w12x E. Fitzer, M. Heine, in: A.R. Bunsell ŽEd.., Fibre Reinforcements for Composite Materials, Elsevier, Amsterdam, 1988. w13x J.F. Moulder, W.F. Stickle, P.E. Sobol, K.D. Bomben, Handbook of X-ray Photoelectron Spectroscopy, Physical Electronics, Eden Prairie, 1995. w14x A. Proctor, P.M.A. Sherwood, J. Electron Spectrosc. Relat. Phenom. 27 Ž1982. 39. w15x T.T.P. Cheung, J. Appl. Phys. 55 Ž1984. 1388. w16x G. Beamson, D. Briggs, High Resolution XPS of Organic Polymers, Wiley, Chichester, 1992. w17x Y. Xie, P.M.A. Sherwood, Appl. Spectrosc. 44 Ž1990. 797, and references therein. w18x H. Buqa, J.O. Besenhard, M. Winter, unpublished results.