Carbon 40 (2002) 145–149

Surface oxides on carbon and their analysis: a critical

assessment
H.P. Boehm*
¨ Munchen
Department Chemie, Universitat ¨ ¨
, Butenandtstr. 5 – 13, 81377 Munchen, Germany
Received 15 March 2001; accepted 15 June 2001

Abstract

The methods for the determination of various types of oxygen surface functions on carbon materials are briefly described,
and their relative advantages and problems that may arise are discussed. Acidimetric titration techniques, IR spectroscopy,
XPS, thermal desorption spectroscopy, and electrokinetic measurements are described.  2002 Elsevier Science Ltd. All
rights reserved.

Keywords: C. Infrared spectroscopy; Temperature-programmed desorption; X-ray photoelectron spectroscopy; D. Surface oxygen complexes

1. Introduction ides can be created by treatment with liquid oxidants, e.g.,

aqueous solutions of H 2 O 2 , NaOCl, (NH 4 ) 2 S 2 O 8 , AgNO 3 ,
Many properties of carbon materials, in particular their H 2 PtCl 6 , etc., at 20–1008C. Oxidation with HNO 3 is often
wetting and adsorption behavior, are decisively influenced used because its oxidizing properties can be controlled by
by chemisorbed oxygen. Oxygen in the surface oxides can concentration and temperature.
be bound in the form of various functional groups which The various aspects of the surface chemistry of carbon
are similar to those known from organic chemistry. This materials have been described in detail [1,7–9]. The
article deals mainly with high surface area carbon materi- present review discusses the most frequently used methods
als which consist predominantly of sp 2 -hybridized carbon for the characterization of surface oxides.
atoms (non-graphitized, turbostratic carbons). The surface
of such carbons is heterogeneous, it consists of the faces of
graphene sheets and of edges of such layers. The edge sites 2. Titration methods
are much more reactive than the atoms in the interior of
the graphene sheets, and chemisorbed foreign elements, in 2.1. Titration of acidic surface functions
particular oxygen, are predominantly located on the edges.
The surface oxides decompose to CO 2 and CO on The surface oxides on a carbon can have acidic as well
heating to high temperatures. Highly reactive sites remain as basic properties and can be conveniently determined by
on the carbon surface which have free-radical character to titration methods. Basic surface character and anion ex-
some relatively small extent [1,2]. After cooling to room change properties are found when a carbon surface,
temperature, they can react with oxygen (air) or water cleaned at ca. 9508C or higher in vacuo or under an inert
vapor, giving new surface oxides. A continuous, very slow gas is exposed to air and aqueous acid (or water) after
oxidation of the surface occurs after a first, rapid cooling to room temperature. Oxygen and acid (or water)
chemisorption of oxygen. The presence of water is essen- are chemisorbed at the same time [9–11]. Surface oxides
tial for this ‘aging’ of the carbons [3,4]. Aging is reduced created with oxygen at elevated temperatures (or by aging)
after prior chenmisorption of hydrogen, e.g., at 9508C or with liquid oxidants are acidic in character and cause
[5,6]. Much more oxygen is chemisorbed at elevated cation exchange properties. Acidic and basic surface sites
temperatures, e.g., 300–4208C. Alternatively, surface ox- coexist usually, but the concentration of basic sites de-
creases with increasing acidic character of the surface.
*Fax: 149-89-2180-7492. The acidic surface properties are caused by the presence
E-mail address: hpb@cup.uni-muenchen.de (H.P. Boehm). of carboxyl groups (also in the form of their cyclic

0008-6223 / 02 / $ – see front matter  2002 Elsevier Science Ltd. All rights reserved.
PII: S0008-6223( 01 )00165-8
146 H.P. Boehm / Carbon 40 (2002) 145 – 149

Fig. 1. A few possible surface groups.

anhydrides), lactones or lactols (see Fig. 1), and hydroxyl constants (Fig. 2). They agree quite well with the titration
groups of phenolic character. These groups differ in their data [13,14]. Such determinations are very time-consuming
acidities and can be differentiated by neutralization with since equilibration in direct titration is very slow [13–15].
0.05 N solutions of NaHCO 3 , Na 2 CO 3 and NaOH, respec- This may be caused by slow diffusion in narrow pores, if
tively. The groups have been identified by other chemical present, and by slow hydrolytic ring opening of carboxylic
methods used in organic group analysis [9,11,12]. Still anhydrides and lactones. The method is limited to a range
higher base uptake than with NaOH is observed with of pKa (or pH) values between 3.5 and 10.5 because of the
alcoholic 0.1 N sodium ethoxide; this can be explained by buffering effect of water at very high or very low pH
the presence of reactive carbonyl groups which form the values. Differences in the number and positions of peaks in
sodium salt of a hemiacetal, =C(OEt)(O 2 Na 1 ). In a the distribution curve have been observed for differently
simple way, the carbons are agitated with an excess of the pretreated carbons.
bases, and the excess is determined by back titration after
equilibration (see Refs. [11,12] for experimental details). 2.2. Titration of basic surface sites
The acidity of a given functional group depends on its
chemical environment, i.e., the size and shape of the While the nature of the acidic surface sites is quite well
polyaromatic layers, the presence and position of other understood, the origin of surface basicity is still under
substituents, and the charge of neighboring dissociated discussion. Continuous titration showed the existence of
groups. However, the differences in acidity of the various three peaks in the pK distribution curves [16]. One reason
types of functional groups seem to be sufficiently large to for a basic behavior of carbon surfaces may be the p
allow differentiation by the simple titration method, e.g., basicity of the exposed graphene layers [17–19]. However,
the difference between NaOH and Na 2 CO 3 consumption this basicity is relatively weak. The chemisorption of
corresponds to the weakly acidic phenolic groups. It has oxygen together with acid suggests that the basicity may
been shown by careful, continuous titration with alkali that be due to oxygen functional groups, and the existence of
several peaks appear in the distribution curve of acidity pyrone-type structures on the edges of the polyaromatic

Fig. 2. Distribution of acidity constants for an activated carbon oxidized with nitric acid (curve taken from Ref. [13], Fig. 8).
 H.P. Boehm / Carbon 40 (2002) 145 – 149 147

C=O vibrations of quinones or isolated carbonyl groups

[24–28]. Peaks at 1000–1300 cm 21 are ascribed to C–O
single bonds [24,25]. Most carbons contain hydrogen, and
band of C–H stretch and wagging vibrations are observed.

4. X-ray photoelectron spectroscopy

In X-ray photoelectron spectroscopy (XPS, also ESCA),

Fig. 3. Pyrone-type structure. core electrons are excited by X-ray irradiation to leave the
atoms. Their binding energy (b.e.) is derived from the
measured kinetic energies. XPS is surface-sensitive since
layers has been suggested [20]. However, it has been the escape depth of the photoelectrons amounts to only a
objected that g-pyrone is a much too weak base to account few atomic layers. The b.e. depends on the atomic species
for the observed strong peak at a pKa of |8.5 of the but is also affected by the shielding of the nuclear charge
conjugate Brønsted acid [16]. Support for the hypothesis of which is lowered or raised by bonding of the atom to more
pyrone-type structures came from theoretical calculations electronegative or electropositive atoms, respectively. The
which showed that the base strength increases strongly differences in b.e. for various binding states are quite small
when the carbonyl group and the ring oxygen of a pyrone- compared to the line width, especially with electronegative
type structure are distributed on polycyclic aromatic elements such as oxygen. A deconvolution of overlapping
compounds, e.g., phenanthrene, as shown in Fig. 3 [21,22]. peaks is necessary, therefore. However, the results of the
Such structures may have basicities even stronger than that curve fitting are influenced to some extent by the some-
of pyridine [21]. what arbitrary inputs for the number, shape and width of
Most carbons contain more oxygen than can be ac- the peaks. Although modern instruments provide a suffi-
counted for by the observed functional groups. The cient resolution also for O1s electrons [15,29], it is more
difference may be due to ether-type oxygen or to carbonyl convenient to measure the C1s signal. Carbon atoms differ
groups that do not react with NaOC 2 H 5 . in their b.e. depending on whether they are linked to one O
atom by a single bond (phenols and ethers), a double bond
(carbonyl groups), or two oxygen atoms (carboxyl groups,
3. Infrared spectroscopy lactones). The corresponding signals appear as satellites on
the high-b.e. side of the main C1s peak of the carbons
Infrared spectroscopy became very useful in analysis of (Fig. 4). The observed ranges of b.e. are listed e.g., in
the functional groups when the limitations due to the high Refs. [15,27,29].
absorbance of carbon were overcome by the development XPS requires calibration since charging of the sample
of FTIR spectroscopy. It is often used in combination with will influence the kinetic energies. Usually, the main C1s
diffuse reflectance spectroscopy (DRIFTS) or with total peak of the carbon samples is taken with a b.e. of 284.6 eV
reflection spectroscopy [23–28]. assigned to it. However, values differing by a few tenths of
The assignment of the absorption bands is based on an eV have been also used (see Ref. [29]). It has been
experience with molecular organic compounds. However, suggested to list rather the shifts relative to the main
there is often disagreement in their assignments. For signals instead of the absolute binding energies [29,31].
instance, an intensive band near 1600 cm 21 has been Alternatively, gold has been sputtered onto the samples
explained by stretching vibrations of aromatic C=C bonds (84.0 eV b.e. for Au 4f 7 / 2 ). It is an advantage of XPS that
which are polarized by oxygen atoms bound near one of the relative surface concentrations of the various species
the C atoms. An alternative assignment to hydrogen- can be estimated from the peak sizes. The results can be
bonded, highly conjugated carbonyl groups was only misleading, however, with porous samples when the
recently refuted [24]. exterior surface is more strongly oxidized by aging.
The spectra of apparently similarly pretreated samples
show often differences, especially in the C=O bands. Free
carboxylic acids absorb near 1700 cm 21 [25], while cyclic 5. Thermal desorption spectroscopy
anhydrides give rise to peaks at 1780 and 1840 cm 21 (sh)
[26]. The anhydrides can be hydrolyzed to the free acids; While carboxyl groups evolve CO 2 at relatively low
the reaction can be reversible on heating. The evidence for temperatures, other groups are thermally more stable.
cyclic lactones is not as conclusive. They are described to Therefore, thermal desorption spectroscopy (TDS), also
give a single peak at 1740 cm 21 [27] or 1760 cm 21 [28], called temperature-programmed desorption (TPD), is used
and a clear distinction from cyclic anhydrides is not for the study of surface oxides. The samples are heated in a
possible. Problems arise also in the assignment of bands to vacuum or in a helium stream with a constant heating rate
148 H.P. Boehm / Carbon 40 (2002) 145 – 149

Fig. 4. Typical C1s XPS spectrum of oxidized carbon fibers: (I) phenols, (II) carbonyl groups, (III) carboxyl groups, (IV) plasmon peak
(after Ref. [30], Fig. 3).

(often 10 K / min), and the evolved gases, H 2 O, CO 2 , CO provide a good overview of the sample’s surface prop-
and H 2 , are determined, mostly by use of a quadrupole erties.
mass spectrometer. The peaks shift to higher temperatures
with increasing heating rate. They are usually very broad,
often with pronounced tailing, and there is considerable 6. Electrokinetic measurements
overlapping. A deconvolution of the spectra into separate
peaks has been described [32]. Techniques to achieve A carbon surface of acidic or basic character is sur-
better resolution and to determine the heat of activation for rounded in aqueous suspension by a diffuse cloud of
the decomposition reactions have been described recently dissociated H 1 or OH 2 ions, respectively, and pH values
[33]. of ,7 or .7, respectively, are measured in such suspen-
It is generally assumed that each type of surface group sions. The pH returns to near neutral, however, after
decomposes to a defined product, e.g., that CO 2 derives sedimentation of the carbon particles when the carbon
from carboxyl groups and CO from carbonyl and hydroxyl sample is electrolyte-free and purified water is used [17].
groups and ether-type oxygen. A good correlation has been In the presence of an electrolyte, ion-exchange results in a
found of the NaOH consumption with the CO 2 -forming permanent pH change. The surface charge depends on the
complexes on the surface of an activated carbon [34]. The pH of the surrounding electrolyte. There is a pH value,
results of TDS are not always unambiguous, however. Two called the ‘point of zero charge’ (PZC) at which the net
adjacent carboxyl groups may be first dehydroxylated to surface charge is zero. The ZPC can be easily determined
the cyclic anhydride which, in turn, decomposes to CO by a method called ‘mass titration’ [36]. It is based on the
plus CO 2 . This decomposition occurs at higher tempera- fact that the pH of an electrolyte changes in the direction
ture than that of free carboxyl groups [32,35]. The of the ZPC on contact with a solid powder.
degradation of lactols will probably also produce both The charged particles move in an applied electric field
gases. A cyclic lactone can either give one CO 2 molecule (electrophoresis). A thin water layer, containing a part of
or two CO molecules, with both reactions possibly running the diffuse cloud of dissociated H 1 or OH 2 ions, adheres
concurrently. Furthermore, secondary reactions cannot be to the particles and moves with it. The charge and potential
excluded. Diffusion of the evolved gases is rather slow in at its boundary determine the electrokinetic phenomena.
narrow pores, and CO molecules may react to CO 2 with The electrokinetic potential (or z-potential) can be calcu-
surface-bound oxygen, or CO 2 molecules hitting the pore lated from the measured electrophoretic mobilities. The pH
walls may form two CO molecules. Freshly created active of zero z-potential is the ‘isoelectric point’ (IEP). It is not
surface sites or free radical sites can facilitate such identical with the ZPC, but usually not very far from it
reactions. with non-porous carbon materials. The observed ZPC
Carbons oxidized with liquid oxidants usually contain range from pH |2 to pH |10.5. With porous carbons,
relatively more carboxyl groups than O 2 -oxidized samples. however, the IEP values are often considerably lower than
Part of these carboxyl groups decompose already at the ZPC because the electrokinetic behavior is determined
temperatures significantly below 3008C. by the charge on the external surface of the carbon
It is not surprising that the TPD spectra published in the particles which is usually oxidized by aging [37]. The PZC
literature often differ in details, since they are influenced is determined, in contrast, by the much larger internal
by the pore structure of the carbons and experimental surface of the pore walls which are oxidized much more
parameters such as the heating rate. Nevertheless, they can slowly since diffusion of oxygen in narrow pores is very
 H.P. Boehm / Carbon 40 (2002) 145 – 149 149

slow. Thus, activated carbons with a predominantly basic ´ y Leon

[18] Leon ´ CA, Solar JM, Calemma V, Radovic LR.
surface may show an IEP in the acidic range. Carbon 1992;30(5):797–811.
[19] Montes-Moran ´ MA, Menendez
´ ´
JA, Fuente E, Suarez D. J
Phys Chem B 1998;102(29):5595–601.
[20] Voll M, Boehm HP. Carbon 1971;9(4):481–8.
References
´
[21] Suarez ´
D, Menendez JA, Fuente E, Montes-Moran ´ MA.
Langmuir 1999;15(11):3897–904.
´ y Leon
[1] Leon ´ CA, Radovic LR. In: Thrower PA, editor, ´
[22] Suarez ´
D, Menendez JA, Fuente E, Montes-Moran ´ MA.
Chemistry and physics of carbon, vol. 24, New York: Marcel Angew Chem, Int Ed Engl 2000;39(7):1376–9.
Dekker, 1994, pp. 213–310. [23] Sellitti C, Koenig JL, Ishida H. Carbon 1990;28(1):221–8.
[2] Lewis IC, Singer LS. In: Walker PL, Thrower PA, editors, [24] Fanning PE, Vannice MA. Carbon 1993;31(5):721–30.
Chemistry and physics of carbon, vol. 17, New York: Marcel [25] Zawadzki J. In: Thrower PA, editor, Chemistry and physics
Dekker, 1981, pp. 1–88. of carbon, vol. 21, New York: Marcel Dekker, 1989, pp.
[3] Billinge BHM, Docherty JB, Bevan MJ. Carbon 147–380.
1984;22(1):83–9. [26] Meldrum BJ, Rochester CH. J Chem Soc, Faraday Trans
´ F, Rivera-Utrilla, Joly JP, Moreno-Castilla
[4] Carrasco-Marın 1990;86(10):1881–4.
C. J Chem Soc, Faraday Trans 1996;92(15):2779–82. ´
[27] Moreno-Castilla C, Lopez-Ramon ´ MV, Carrasco-Marın ´ F.
[5] Verma SK, Walker Jr. PL. Carbon 1992;30(6):837–44. Carbon 2000;38(14):1995–2001.
´
[6] Menendez JA, Phillips J, Xia B, Radovic LR. Langmuir [28] Zhuang Q-L, Kyotani T, Tomita A. Energy Fuels
1996;12(18):4404–10. 1994;8(3):714–8.
[7] Bansal RC, Donnet JB, Stoeckli F. Active carbon. New [29] Papirer E, Lacroix R, Donnet JB. Carbon 1994;32(7):1341–
York: Dekker. 1988:27–118, 259–333. 58.
[8] Donnet JB, Bansal RC, Wang MJ. In: Carbon black, New [30] Yue ZR, Jiang W, Wang L, Garner SD, Pittman CU. Carbon
York: Dekker, 1993. 1999;37(11):1785–96.
[9] Boehm HP. In: Delhaes ¨ P, editor, Graphite and precursors, [31] Nanse´ G, Papirer E, Fioux P, Moguet F, Tressaud A. Carbon
Amsterdam: Gordon and Breach, 2001, pp. 141–78. 1997;35(2):175–94.
[10] Voll M, Boehm HP. Carbon 1970;8(6):741–52. [32] Figureido JL, Pereira MFR, Freitas MMA, Orfao ´ ˜ JJM.
[11] Boehm HP. Carbon 1994;32(5):759–69. Carbon 1998;37(9):1379–89.
[12] Boehm HP, Diehl E, Heck W, Sappok R. Angew Chem, Int [33] Haydar S, Moreno-Castilla C, Ferro-Garcıa ´ MA, Carrasco-
Ed Engl 1964;3(17):669–78. Marın´ F, Rivera-Utrilla J et al. Carbon 2000;38(9):1297–
[13] Contescu A, Contescu C, Putyera K, Schwarz JA. Carbon 308.
1997;35(1):83–94. [34] Otake Y, Jenkins RG. Carbon 1993;31(1):109–21.
[14] Bandosz TJ, Jagiello J, Contescu C, Schwarz JA. Carbon [35] Moreno-Castilla C, Carrasco-Marın ´ F, Maldonado-Hodar
´ FJ,
1993;31(7):1193–202. Rivera-Utrilla J. Carbon 1998;36(1):145–51.
[15] Biniak S, Szymanski J, Siedlewski J, Sviatkowski A. Carbon [36] Noh JS, Schwarz JA. Carbon 1990;28(5):675–82.
1997;35(12):1799–810. ´
[37] Menendez ´ ´
JA, Illan-Gomez ´ y Leon
MJ, Leon ´ CA, Radovic
[16] Contescu A, Vass M, Contescu C, Putyera K, Schwarz JA. LR. Carbon 1995;33(11):1655–9.
Carbon 1998;36(3):247–58.
[17] Boehm HP, Voll M. Carbon 1970;8(2):227–40.