This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
(3.8b)
Equation (3.8) shows that the equilibrium enthalpy can be calculated from experimental data, i.e., if the dependence of the isotherm on temperature at constant spreading pressure is known. However the use ofEqns (3.8a) and (3.8b) is cumbersome because it requires the previous calculation of the spreading pressure instead of the use of raw experimental data, i.e., n(a) vs p. Furthermore, the molar quantities appearing in Eqn (3.7) are not the natural variables for adsorption systems. In preference to molar quantities, partial molar entropy and internal energy are generally used; these quantities measure the changes in these properties when an infinitesimal change occurs in the number of adsorbed moles at constant temperature, pressure, and area. To be able to relate these quantities to experimental measurements, differentiation of the chemical potential of the adsorbed phase (in Eqn (3.7)) and rearrangement finally leads to (3.9) This expression gives the definition ofthe enthalpy known as isosteric enthalpy of adsorption. Equation (3.9) can be simplified to
T (V(g) -
(3.10) v(a))
3.2 Classical Thermodynamics
57
Again, as in the case of Eqn (3.8), if the adsorbed phase is assumed to be liquid-like, Eqn (3.10) reduces to
(
dlnp ) dT n,A
(3.11)
There is at least one other enthalpy related to the experimental data. This enthalpy is obtained in a calorimetric experiment under adiabatic conditions. The experiment consists of adding gas, in a reversible manner, to the calorimeter containing the adsorbent. An alternative process could be considered as a way to simplify the problem. Instead of adding gas to the system, imagine that the adsorbed molecules are transferred from the gas phase to the adsorbed phase by the action of a piston that changes the gas phase volume by an amount d V(g) . Assuming that the area of the adsorbate is unchanged during the process, that the adsorbed phase is liquid-like, and that the gas phase is ideal, it is possible to derive the expression _ qad -
qst
+ V (g)
(
dp ) d (a) n
(3.12) ad
This relationship shows that it is possible to calculate the isosteric enthalpy of adsorption from calorimetric experiments. In summary, it has been shown how the enthalpies of adsorption are obtained either calorimetrically or from the dependence of the isotherms on temperature. Although the definitions given above for the different enthalpies of adsorption are rigorous, it is necessary to show that they exhibit the same properties as the enthalpy of vaporization, i.e., the heat necessary to vaporize one mole ofliquid at constant pressure and using a reversible and isothermal process. Moreover, only the isosteric enthalpy is related in a simple way to the heat required in a process that is reversible and isothermic. Suppose that moles of adsorbate are transferred to the gas phase at constant temperature and pressure. According to the first law of thermodynamics:
an
(3.13) If the area of the solid is kept constant, this expression leads to
(oQ) A
ou(a)) = (On(a)
On(a) + T,p,A
OV(g)) +p ( - on(g) On(g)
(OU(g))
-On(g)
on(g)
+ p (ov(a)) -On(a)
on(a) T,p,A
(3.14)
Using the definitions of mean molar and partial molar quantities, Eqn (3.14) becomes (3.15)
Chapter 3 Energetics of Gas Adsorption by Carbons: Thermodynamic Quantities
58
Since Bn(g)
= -on(a) = On, Eqn (3.15)
can be written as (3.16)
According to Eqn (3.9) this quantity is equal to the isosteric enthalpy, qst' which turns out to be equal to the heat per mole evolved in the reversible transfer of an infinitesimal amount of adsorbate from the adsorbed phase to the gas phase at constant temperature, pressure, and area. Up to this point the essential equations have been presented. Now, it is possible to analyze the work carried out in connection with the classical thermodynamic approach. The first systematic study of a thermodynamic adsorption quantity was perhaps the work done by de Boer and coworkers [10] on the determination, interpretation and significance of the enthalpy and entropy of adsorption. Their papers analyzed almost all aspects of the experimental determination of the entropy and how to interpret the values obtained in terms of two extreme models, i.e., those of mobile and localized adsorption, which today have lost much of their usefulness. To catalog the behavior of the adsorbed film as localized or mobile is a very simplistic solution and it has been demonstrated [9] that in most cases the adsorbed film is neither completely localized nor completely mobile. This approach also is somehow outdated because numerical simulations provide a better microscopic interpretation of the system's behavior. Fomkin et al. [11] have reported a slightly different treatment in which they use Eqn (3.11) to calculate the isosteric heat of adsorption of perfluoropropane adsorbed on PAC (powdered activated carbon) microporous carbon. Agarwal et al. [12] determined the entropy of the adsorbed phase for methane, ethane, ethylene, propane, carbon dioxide, and nitrogen adsorbed at high pressures on activated carbon. Bottani et al. [13] also employed the classical approach to obtain the entropy of the adsorbed phase for N 2 and CO 2 adsorbed on "graphitized" carbon blacks. The authors discussed several problems regarding the precision of the obtained values using Eqn (3.11) or equivalent equations, and how they could be employed to characterize the surface of carbonaceous materials. More recently, Sircar et al. [14] employed the Gibbsian Surface Excess model to describe the multicomponent adsorption of gas mixtures. They also showed that this model for multicomponent adsorption could define unambiguously the isosteric heats of adsorption for the components of a gas mixture. These variables can be experimentally determined using multicomponent differential calorimetry and then be used to describe the nonisothermal behavior of practical adsorbents. Mezzasalma [15] employed a condition of maximum irreversible entropy production in the framework of a variational procedure where the isotherm equations are represented by a convergent sequence of ordinary functions. Milewska-Duda et al. [16] employed the thermodynamic approach described above to derive an adsorption isotherm, similar to the BET equation, which
3.3 Statistical Mechanics
59
can describe adsorption in microporous structures provided that restrictions for pore capacity are taken into account. Asnin et al. [17] demonstrated that the classical thermodynamic approach does not contradict the molecular statistical theory and that it yields equations that are more general. Based on Eqn (3.4) and similar expressions for the internal energy, they analyzed the particular case of the Freundlich adsorption isotherm. With data obtained from Kr adsorption on high-modulus carbon fibers, Drzal et al. [18] determined the isosteric heat of adsorption and the entropy of the adsorbed phase and demonstrated that such fibers, which undergo a high-temperature graphitization treatment, possess a very homogeneous surface very similar to that of the basal plane of graphite. Sircar [19] has presented a thermodynamic treatment ofgas mixture adsorption on heterogeneous adsorbents with particular emphasis on the estimation of the isosteric heat of adsorption. He stated that the isosteric heat of adsorption on an energetically heterogeneous adsorbent could vary substantially depending on the fractional loadings of the adsorbates, which, in turn, depend on the equilibrium gas-phase pressure, temperature, and composition. Myers [20] has developed thermodynamic equations for adsorption of multicomponent gas mixtures on microporous adsorbents based on the principles of solution thermodynamics. He argued that the conventional spreading pressure and surface variables, which describe bidimensional films, must be abandoned for adsorption in micropores, in which spreading pressure cannot be measured experimentally or calculated from intermolecular forces. Li et al. [21] recently reviewed the progress made in predicting the equilibria of multicomponent mixture adsorption. They discussed the problems encountered in applying theories developed for subcritical mixtures to supercritical gases. In a recent paper, Chiang et al. [22] reported values of the free energy, enthalpy, and entropy of adsorption of volatile organic compounds (exemplified by benzene and methylethylketone) on seven samples of activated carbon. The starting point for their development was Eqn (3.11) for the isosteric heat of adsorption. Linders et al. [23] determined adsorption heats from the adsorption equilibrium constant and found that these values agree quite well with those obtained from uptake experiments using the integrated form ofEqn (3.11). They analyzed the experimental data obtained for n-butane adsorbed on two commercial activated carbons (Kureha and Sorbonorit B3) and for hexafluoropropylene adsorbed on activated carbon.
3.3
STATISTICAL MECHANICS
The statistical mechanics formalism is probably the most efficient way to connect molecular models with experimental data. We present here a brief summary of the most important equations used for numerical simulations. Of all the statistical ensembles that can be employed, the canonical and grand canonical
60
Chapter 3 Energetics of Gas Adsorption by Carbons: Thermodynamic Quantities
are the most popular. We also restrict our treatment to classical statistical thermodynamics. Thus no quantum effects are taken into account. The probability that molecules 1, 2, 3, ... , N are in the volume elements dr l' dr 2' • • • , dr N located at r l' r 2' •• • , r N is given by the Boltzmann expression (3.17)
where
zt)
is the normalization factor given by (3.18)
Here V is the volume available to the gas molecules and U(r 1 , r 2 , ••• , r N ) is the potential energy of the N molecules. Thus U can be taken as the sum of two terms, gas-solid (Ugs ) and gas-gas (Ugg ) interaction energies: N
L
U(r1,r2, ... ,rN)=LUgs(r;)+
Ugg(rij)
(3.19)
l~i<j~N
i=l
It must be pointed out that this expression implicitly contains terms that depend on the orientation of the molecule with respect to the surface and the orientation of a given molecule with respect to its neighbors when the molecules are nonspherical. Equation (3.19) assumes that the potentials are additive and pairwise; since it does not include three-body or higher terms, this must be considered as an effective potential [8]. A system that is constrained to have a constant number of molecules, volume, and temperature constitutes a canonical ensemble. The thermodynamic properties of the system can be calculated from the corresponding partition function, Q(N, V, 7) [24]. For the adsorbed phase the partition function can be written as z(a) N Q( N " V T) -- N!A3N
(3.20)
where A is a factor that includes the kinetic properties of the molecule. Equation (3.20) shows that the partition function can be written as the product of a configurational factor and a kinetic or nonconfigurational factor, A. This greatly simplifies the application of this approach to the theory of physical adsorption [8]. The main assumption implicitly contained in Eqn (3.20) is that the structural properties of the molecules are independent of the intermolecular interactions in all the important configurations of the system. Even though there is evidence to suggest that this is not strictly true, it is possible to derive the appropriate expressions to calculate the extent of the changes in those properties
61
3.3 Statistical Mechanics
when the molecule is adsorbed. The thermodynamic properties of the system can be calculated from the following expressions:
(3.21)
A=-kTlnQ alnQ )
U=-k a(l/T) (
Q) aln-kT ( av(a)
(3.22) N,v(a),A
(3.23)
P-
N,T,A
Q
A,. (aln o/-kT - -)
aA
(3.24) NT vCa)
where A is the Helmholtz free energy, U is the total energy, p is the pressure, and time) or the ensemble-averaged value of a property from MC) « A > ens') The goal of a computer simulation is to generate enough states or configurations of the system of interest to be able to evaluate such time or ensemble averages.
Chapter 4 Monte Carlo and Molecular Dynamics
9°
Figure 4.4 Shapshots of 90 molecules of adsorbed CO 2 at 195 K. The black bands indicate the central carbon atoms in each molecule. The upper panel is a side view of all 90 molecules and the lower panel is a top view of the 78 molecules in the first layer only.
4.2.4.1 Ensemble averaging In a MC simulation, the ensemble average of a property is determined by summing the value of the property in each configuration and dividing by the total number of configurations [7]. In general,
A obs = (A) ens A obs = (A)ens
= (A (f (r))) = -
1
Tobs
Tobs
= (A(f(r))) =
LA (f (r))
(4.8)
T=l
1
Tobs
-
LA(f(T))
'T obs T=l
where A obs is the observed value of a property, ens is the ensemble-averaged value of the property, fer) is the generalized coordinates (positions only) T is the index over state points, and 'Tobs is the total number of states or points generated by the MC prescription.
91
4.2 Overview of Computer Simulations
For example, in the simulation of Ar and CO 2 on C 60 fullerene by MartinezAlonso et al. [21] , the simulated isotherm was compared to experimental isotherms. Using a grand canonical ensemble, the isotherm was obtained by calculating the average number of particles in the simulation cell (Nens ) for each fixed value of the chemical potential using the MC method to generate configurations (positions of the particles in the simulation cell). 1
N ens = (N(r)) = -
Tobs
LN(r)
(4.9)
robs T=l
1
N ens = (N(r)) = -
Tobs
LN(r)
robs
1
where N(r) is the total number of particles in the simulation volume in the rth configuration. In addition to the isotherm, the configurational information can be used to obtain local density distributions of the particles on an adsorbing surface. For instance, if the adsorbent is made up of C 60 particles, the density as a function of distance from the surface shows the layers of adsorbate forming and lends insight into the structure of the surface layers as shown in Fig. 4.5. 2.5 , . . - - - - - - - - - - - - - - - - - - - - - - - - - - . ,
2.0
1.5
1.0
0.5
0.0 .....__ __...,..,IIII__...J....__;;;;;L...._...4.__ 1.2 1.4 1.6 1.8 2.0 2.2 ~
__1__
2.4
__L.,;;;===___L__........&_._
2.6
2.8
3.0
___L._---J
3.2
3.4
Z(nm)
Figure 4.5 Simulated densities of Ar at 77.5 K (thick line) and CO 2 at 195.5 K (thin line) adsorbed on an fcc array of C60 molecules approximated as spherical bodies are plotted as a function ofdistance from the solid adsorbent. Adsorption on the outermost layer ofC60 spheres produces three density peaks: one at 1.4 nm in the deep (and strongly interacting) crevices located in the centers ofthe squares formed by four C60 molecules; one at 1.8 nm in the crevices between pairs of neighboring C60's; and one at rv2.1 nm for adsorbed molecules directly over a C60. Peaks at larger distances reflect structure in the adsorbed fluid, with the CO 2 density decreasing to zero after second layer formation at 195.5 K because of its relatively small amount in the simulation box compared to the Ar multilayer densities at 77.5 K.
92
Chapter 4 Monte Carlo and Molecular Dynamics
4.2.4.2 Time averaging In MD simulations, both the positions and velocities of the particles are calculated and because their time dependence is known, both thermodynamic and dynamic properties can be calculated. This is the main advantage of MD as a method for generating configurations. In this case, the time-averaged properties can be evaluated [7]:
A obs = (A)time
lim
1 {tobs
= (A (r(t))) = - - 1n
A (f(t))dt tobs 0 tobs lim 1 A obs = (A)time = (A(r(t)) = - - - A(r(t))dt tobs~oo tobs 0
(4.10)
tobs~oo
f
where A obs is the observed value of a property, < A >time is the time-averaged value of the property, and is the generalized coordinates (positions and momenta) as a function of time. In the recent simulation by Matties and Hentschke [36, 37], the adsorption and melting ofbenzene on graphite was studied via MD simulations. In addition to determining static properties such as the center of mass density distributions and tilt angles as a function of temperature by obtaining time averages, they were also able to obtain dynamic properties such as the surface diffusion constants in the monolayer and the orientational velocity autocorrelation function (OVAF).
ret)
OVAF = Z (r)
= (vxy (t) • vxy (t + r))
(4.11)
This function is a measure of the reorientation of the component of the velocity vector parallel to the surface vxy . It is calculated by choosing a molecule and following its motion as a function of time for a specified time period, averaging the velocity autocorrelation function (the dot product of the 2D vector velocity at the time t, vxy(t) with the velocity at a later time in the trajectory (vxy(t + r)). This average is calculated for all molecules and for many different initial starting times. The determination of this function can be used to understand the motion of the adsorbed benzene molecules on the surface. Negative values of this time-correlation function are the result of a constraining environment representative of a solid structure so it serves as an indicator of the transition from a solid to a liquid monolayer as temperature is increased [36]. Matties and Hentschke [37] are also able to show that even at high coverage (multilayer adsorption) and high temperature there is some semblance of solid order in the adsorbed benzene layers (see Fig. 4.6).
4.2.4.3 Results using different thermodynamic ensembles Historically, simulations using the microcanonical ensemble were among the earliest ones reported. The algorithm is easy to 5mplement both conceptually
93
4.2 Overview of Computer Simulations
0.04
180.......200··+---
0.02 0
E I\J
-0.02 240 ---260 -
-0.04
280 ··41·'-'· _~~_ ...
---------- -. 300 -..
-0.06
320 .-..--
-0.08 -0.10
o
0.2
0.4
0.6
0.8
1.0 T
0.04
1.2
1.4
1.6
1.8
2.0
(pS)
200 -+-220-+-
(b)
0.02
240··.····-
0
E
I\J
-0.02 260--
-0.04
280 ---300··..·•·• 320 .....--·
-0.06 -0.08 -0.10 0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8 2.0
T(pS)
Figure 4.6 Linear velocity time correlation functions are shown for benzene on graphite at coverages of two (a) and three monolayers (b). These curves are averages over all molecules of the product of the molecular velocity vector parallel to the surface at time zero and the value of this vector at time r, normalized to unity at time zero. Numbers at the right denote temperature in Kelvin. Curves for the highest temperatures decay monatomically to zero, as is typical for weakly interacting fluids where there is no significant cage formation. The negative values for Z(r) indicate anticorrelation in the velocity direction, characteristic of motion within a constraining cage.
and technically since the volume (V) (simulation boundaries) and the number of particles (N) remain constant. In addition, the total potential energy (E) is held constant: (N, V, E). Although the total density (N/ V) is constant in a bulk phase, in simulations of surface phenomena the presence of the external potential (adsorbing surface) produces large variations in the density, making this ensemble useful for studying structures of adsorbed phases and energy distributions.
94
Chapter 4 Monte Carlo and Molecular Dynamics
Furthermore, while the total energy is held constant, the energy is distributed between the molecules adsorbed and in the gas phase. Therefore, the distribution of energy between these two phases can be explored and compared to experimental isosteric heats of adsorption. For example, the paper by Steele et al. [38] summarizes the results obtained for N 2 adsorbed on graphite using MD simulations in the microcanonical ensemble. Besides calculating heats of adsorption, they were able to study how the energy of the adsorbate molecules varied along the surface and use this to understand the 2D phase diagram of N 2 on graphite and the role of the quadrupole moment in the formation of the surface structures. However, despite the ease of simulations in this ensemble, it is not easy to devise an experiment where the internal energy is held constant and it is more natural to look for a technique that corresponds to the more common experimental situation of constant temperature. Canonical ensemble (N, V, T) is a variation on the microcanonical ensemble where the temperature is held constant rather than the energy. In the constraint method described by Evans and Morris [39], this is done in a MD simulation by rescaling the velocities to maintain a constant kinetic energy. A more rigorous algorithm developed by Nose [40] and reformulated by Hoover [41] introduces an additional degree of freedom into the simulation that acts as a heat bath. Energy is exchanged with the artificial coordinate and its velocity introduced to maintain a constant temperature, but nevertheless allowing for fluctuations in the kinetic energy. Studies of these techniques show that the Nose algorithm does indeed produce the correct canonical distribution in both position and momentum space. However, the isokinetic scheme has been shown to give correct canonical ensemble averages for properties that depend on coordinates [42]. Kim and Steele [43] used isokinetic MD simulations to study the phase diagram of the methane monolayer on a corrugated graphite surface. They were able to compare their simulation results for the structure of the methane monolayer to the results of neutron scattering experiments and found good agreement. Trajectory plots (plots of positions of the molecules as a function of time) obtained from this simulation at four temperatures are shown in Fig. 4.7. These traj ectories show the transition from a commensurate to an incommensurate solid, even hinting at the possibility of a two-phase coexistence. The study of phase transitions and structure is one of the goals of computer simulations; however, determining the pressure in the canonical and microcanonical ensemble can be difficult. It is done using either thermodynamic integration [44, 45], or the test particle method of Widom [46]. Both of these techniques are computationally intensive, so the development of an ensemble where the pressure can be held constant is desirable. In the isobaric, isothermal ensemble (fixed P, N, T) the pressure is held constant by varying the volume of the simulation cell. Finn and Monson [47] have developed a method for studying adsorption in an isobaric algorithm. Unfortunately, if this method is used in a simulation involving a solid surface one would have to be sure that variations of the volume are accompanied by increases or decreases in the surface area of the adsorbent. This can create significant problems in the potential
4.2 Overview of Computer Simulations
95
:-~:.: ... :~;;T!
I
.- __ _
--\l....~t ~.
••
.
~-::~. ~
~.I~;1 \ • : • • ~ .~~ .
:.~. :r-.v:. ~. ·
. ·:·:·:·:·.·,i~!
~
o~
~
• •
r
I
""
~, ~ ~ ~ t •\ .~ , ••••••••••••••• 'I
• • • • • • • • • • • • ' • • • • • • • 'j
. . . . . . . '! o. . . . . . . . . . I ... .. ..... ..
..................... 'I
..
I'
;1
~
•
•
•
•
•
l J ' •
•
"
•
•
•
•
•
•
•
•
•
•
•
•
•
• •
•
•
•
• •
•
•
lit
•
•
•
•
•
•
•
• •
•
-
•
•
•
•
•
•
•
•
t
•
•
•
•
to
•
•
~~ ~ '.' ,,~~ !oI·f-""'.""""'-' ;
.~ ~~·:-:-:-:t / ~1i:.: ·:·: ~ N .: ~
J.
1
51 K
56K
Figure 4.7 Plot of the in-plane trajectories of adsorbed molecules at four temperatures are shown here for an incommensurate solid methane monolayer adsorbed on graphite. At the lowest temperatures, the figures show large solid patches with a few drifting particles near the patch edges. As T increases, molecules begin to enter different sublattices and to jump from one to another. (Each plot has a duration of 88 ps.) The dots are for molecules that are vibrating over a single site on the graphite and those molecules that shift from one site to another give rise to larger excursions. The trajectory plots at 55 and 56 K seem to indicate the two-phase coexistence. This possibility is supported by simulated methane-methane interaction histograms for these temperatures that show two peaks in the energy distribution, one near the maximum value for solid and one near the maximum value for the liquid. The fact that this behavior extends over a range of T is probably a finite size effect.
description (discontinuities at the boundaries must be avoided) as well as in the determination and interpretation of most structural properties. Another popular method for studying phase transitions involves the use of the grand canonical ensemble (fixed /-L, V, T), which is most commonly implemented using the MC method (gcmc) described in Section 4.2.3.2. Since the development of this technique it has become the method of choice for the determination of structural and thermodynamic properties of most adsorbateadsorbent systems. For example, Bottani et al. have recently presented gcmc simulation results for gases such as N 2 , Ar, and CO 2 adsorbed on carbon nanotubes [48] and on C 60 [49]. In their studies, the authors simulate the adsorption isotherms (number ofa particles as a function of P or /-L), and energy distributions that allow comparison of their simulations with experiments. Furthermore, after
Chapter 4 Monte Carlo and Molecular Dynamics
establishing a firm connection between simulation and experiment, the authors use the density profiles obtained from the simulation to understand where gas molecules adsorb. In studies of bundles of carbon nanotubes, by looking at density distributions they see significant adsorption in the interstitial channels and external surface at low pressures before all interior sites are full even though it is clear that adsorption inside nanotubes is preferred to adsorption outside the tubes (see Fig. 4.8). The Gibbs ensemble is a technique that allows one to study phase equilibria without an interface, by combining two simulations at the same time. In this method originally proposed and developed by Panagiotopoulos [50] two simulation cells are set up: each cell represents one of the two phases in equilibrium with each other. In this algorithm, the total N, V, and T are held constant; however, N and V vary in the separate simulation cells. The acceptance conditions for the various MC moves maintain the same chemical potential (i.e., equilibrium) in the two simulation cells. The method involves the execution of three types of MC trial moves. The first is the displacement of a randomly
(a)
(c)
(b)
35,.----------, o
46
Q
o
41 36 31 ~ 26
21 16 11 6
6 11 1621 2631 3641 46
X
Figure 4.8 Maps of the average density of nitrogen adsorbed in three nanotube bundles. The contours are for constant density in the x, y planes; Le., for an observer looking in the z direction parallel to the pore axes. The pore diameters are (a) 1.37 nm, (b) 1.43 nm, and (c) 0.69nm. The in-plane coordinates x, yare defined so that unit x, y== 0.07, 0.14nm, respectively. The larger blobs show density contours inside the tubes and the smaller ones are for molecules adsorbed in the interstices between the hexagonally packed tubes. The interaction potential for the N 2 is diatomic; thus, the approximate molecular length is 0.1 nm greater than the width which is 0.35 nm. The consequence is that the tube of (c) is too small to admit the N 2 molecules so that the adsorption shown there is essential all interstitial. Also, in (a) and (b), the N 2 appears to lie parallel to the tube axis and is adsorbed on the tube walls. The differences between the (a) and (b) contours are at least partly due to the differences in the numbers of molecules in these systems. These amount to 334 and 199 in (a) and (b).
4.3 Conclusions
97
chosen molecule in each cell to maintain internal equilibrium. The second is a change in the volume of one of the simulation cells that is accompanied by a corresponding volume change in the other such that the total volume remains constant. The third type of move involves the transfer of a randomly chosen particle from one simulation cell to the other (to maintain chemical equilibrium). Panagiotopoulos [51] extended the technique to simulate the coexistence offluids adsorbed in micropores. In this study, he simulated both the coexistence between the pore fluid and the bulk fluid (to obtain the adsorption isotherm) and capillary condensation in cylindrical pores. Since its development this method has been used in studies of both vapor-liquid [31] and liquid-liquid [52] phase transitions in carbon nanopores.
4-3
CONCLUSIONS
MC and MD are versatile techniques that have been shown to be powerful methods of enhancing our understanding of molecular behavior both of carbon surfaces and of the many other solid adsorbents presently in use. Although this chapter has dealt with the basics of computer simulation, there are many areas where simulators have been active that have not been dealt with in the chapter (e.g., see Chapters 5,6,8-10, and 15). In general, MD is used when transport properties are desired and MC, when thermodynamic properties are the subject of interest. An important feature is that a wide variety of experimental systems are encountered in this field, starting with flat, homogeneous, chemically uniform surfaces such as graphite, metals, and single crystals of ionic material. The algorithms initially developed to deal with such cases were soon modified and altered to handle porous materials or materials with impurities and/or imperfections in their exposed surfaces. This has enabled the researcher to dispense with the older theories that by necessity included approximations that inevitably had the possibility of invalidating the results of analyses based on such oversimplified models. It is probably fair to say that many of the advanced algorithms now in use might never have been developed if they were not required for the simulation of relevant adsorption systems. Fortunately, the advances in the analysis of complex adsorption systems have coincided with the impressive improvements in computing power needed to carry out these analyses. There are still new and exciting areas under development. These include the path integral Monte Carlo (PIMC) method where quantum systems interacting with graphite can be studied. Manousakis et al. have used the PIMC technique to study 4He and H 2 films [53-57] on graphite. They are able to simulate the low temperature structural properties, including 2D phase transitions. Johnson et al. have developed a path integral gcmc technique that allows them to calculate adsorption isotherms. Using this method, Wang and Johnson have simulated H 2 and He in carbon slit pores [58] and carbon nanotubes [59]. They have studied
Chapter 4 Monte Carlo and Molecular Dynamics
H 2 storage in graphite nanofibers [60], the feasibility of using carbon nanotubes to separate hydrogen isotopes [61] and phase behavior of H 2 and He isotopes in nanotubes [62]. Another exciting new direction that is developing is a technique known as reverse MC [63] (also see Chapter 5 of this book). In this method, rather than performing a simulation to gather configurations on an assumed solid, MC moves are made on the atomic configuration of a simple model adsorbent in an attempt to move from the original configuration to a configuration, which agrees with a previously chosen experimental property of the adsorbent such as the structure factor. The generation of carbon adsorbents whose structures match the available experimental data has been investigated by Gubbins and coworkers [64-67] using this method. Reference [66] gives a brief review of previous efforts to deal with the structural problem for porous carbons. Once model adsorbents are generated using the constrained Me technique, Gubbins et al. perform standard gcmc simulations ofN2 adsorbed on their model systems to determine pore size distributions, porosity and heats of adsorption of the model surfaces. Although this technique is quite computer-intensive, the resulting structures appear to be good representations of porous carbon as indicated by the agreement with experiment of simulations of the adsorption of simple gases in the model samples.
REFERENCES
1. Rowley, L.A., Nicholson, D., and Parsonage, N.G. (1976). Grand ensemble Monte Carlo studies of physical adsorption I. Results for multilayer adsorption of 12-6 argon in the field of a plane homogeneous solid. Mol. Phys., 31, 365-87. 2. Rowley, L.A., Nicholson, D., and Parsonage, N.G. (1976). Grand ensemble Monte Carlo studies of physical adsorption II. The structure of the adsorbate. Critique of theories of multilayer adsorption for 12-6 argon on a plane homogeneous solid. Mol. Phys., 31, 389-407. 3. Rowley, L.A., Nicholson, D., and Parsonage, N.G. (1978). Long-range corrections to grand canonical ensemble Monte Carlo calculations for adsorption systems. J. Comput. Phys., 26, 66-79. 4. Talbot, J., Tildesley, D.J., and Steele, W.A. (1984). A molecular dynamics simulation of nitrogen adsorbed on graphite. Mol. Phys., 51, 1331-56. 5. Talbot, J., Tildesley, D.J., and Steele, W.A. (1986). Molecular dynamics simulation of fluid N 2 adsorbed on a graphite surface. Faraday Disc. Chem. Soc., 80, 91-105. 6. Talbot, J., Tildesley, D.J., and Steele, W.A. (1986). A molecular dynamics simulation of the uniaxial phase of N 2 adsorbed on graphite. Suif. Sci., 169, 71-90. 7. Allen, M.P. and Tildesley, D.J. (1987). Computer Simulation of Liquids. Oxford University Press.
References
99
8. Frenkel, D. and Smit, B. (2002). Understanding Molecular Simulation, From Algorithms to Applications, 2nd edna Academic Press. 9. Haile, J.M. (1992). Molecular Dynamics Simulations: Elementary Methods. J. Wiley and Sons. 10. Balbuena, P.B. and Seminario, J.M. (eds) (1999). Molecular Dynamics, From Classical to Quantum Methods. Elsevier. 11. Bruch L.W., Cole, M.W., and Zaremba, E. (1997). Physical Adsorption: Forces and Phenomenon. Clarendon. 12. Steele, W.A. (1973).The physical interaction of gases with crystalline solids. I. Gas-solid energies and properties of isolated adsorbed atoms. Surf. Sci., 36, 317-52. 13. Steele, W.A. (1978).The interaction of rare gas atoms with graphitized carbon black.]. Phys. Chem., 82, 817-21. 14. Watts, R.O. and McGee, I.J. (1976). Liquid State Chemical Physics. Wiley. 15. Maitland, G.C., Rigby, M., Smith, E.B., and Wakeham, W.A. (1987). Intermolecular Forces: Their Origin and Determination, Tables A3.1 and A3.2. Clarendon, pp.565-66. 16. Steele, W.A. (1974). The Interaction of Gases with Solid Surfaces. Pergamon Press. 17. Murthy, C.S., Singer, K., Klein, M.L., and McDonald, I.R. (1980). Pairwise additive effective potentials for nitrogen. Mol. Phys., 41, 1387-99. 18. Bojan, MJ., van Slooten, R., and Steele, W.A. (1992). Computer simulation studies of the storage of methane in microporous carbons. Separation Sci. Technol., 27, 1837-56. 19. Vernov, A.V. and Steele, W.A. (1986). Computer simulation of the multilayer adsorption of fluid N 2 on graphite. Langmuir, 2, 219-27. 20. Bojan, M.J., Vernov, A.V., and Steele, W.A. (1992). Simulation studies ofadsorption in rough-walled cylindrical pores. Langmuir, 8, 901-8. 21. Martinez-Alonso, A. Tascon, J.M.D., and Bottani, EJ. (2001). Physical adsorption of argon and CO 2 on C 60 fullerene.]. Phys. Chem. B, 105, 135-9. 22. Stan, G., Bojan, M.J., Curtarolo, S., et al. (2000). Uptake of gases in bundles of carbon nanotubes. Phys. Rev. B, 62, 2173-80. 23. Calbi, M.M., Gatica, S.M., Bojan, MJ., and Cole, M.W. (2001). Phases of neon, xenon, and methane adsorbed on nanotube bundles.]. Chem. Phys., 21, 9975-81. 24. Maddox, M.W. and Gubbins, K.E. (1995). Molecular simulation offluid adsorption in buckytubes. Langmuir, 11, 3988-96. 25. Kim, H.Y. and Cole, M.W. (1987). Three-body contribution to the adsorption potential of atoms on graphite. Phys. Rev. B, 35, 3990-4. 26. Roth, M.W. (1998). Bond-orientational structure and melting signature in krypton physisorbed onto graphite at complete coverage. Phys. Rev. B, 57, 12520-9. 27. Bojan, M. J., and Steele, W. A. (1993). Computer simulation studies of diffusion in physisorbed monolayers. Mater. Res. Soc. Symp. Proc., 290, 127-34. 28. Kolafa. J. (1988). On optimization of Monte Carlo simulations. Mol. Phys., 63, 559-79. 29. Norman, G.E. and Filinov, V.S. (1969). Investigations of phase transitions by a Monte Carlo method. High Temp. (USSR), 7, 216-22. 30. Bojan, MJ., Bakaev, V.A., and Steele, W.A. (1999). Smart Monte Carlo algorithm for the adsorption of molecules at a surface. Mol. Simul., 23, 191-201. 31. Jiang, S. and Gubbins, K. E. (1995).Vapor-liquid equilibria in two-dimensional Lennard-Jones fluids: unperturbed and substrate-mediated films. Mol. Phys., 86, 599-612.
100
Chapter 4 Monte Carlo and Molecular Dynamics
32. Cracknell, R.F., Nicholson, D., Parsonage, N.G., and Evan, H. (1990). Rotational insertion bias: a novel method for simulating dense phases of structured particles, with particular application to water, Mol. Phys., 71, 931-43. 33. Ulberg, D.E. and Gubbins, K.E. (1994). Monte Carlo implementation on the Connection Machine 2; water in graphite pores. Mol. Simul., 13,205-19. 34. Bottani, E., Bakaev, V.A., and Steele, W.A. (1994). A simulation/experimental study of the thermodynamics properties of CO 2 on graphite. Chem. Eng. Sci., 49, 2931-9. 35. Avgul N.N. and Kiselev A.V. (1970). In Chemistry and Physics of Carbon, Vol. 6, p. 65 (P. Walker, ed.). Marcel Dekker. 36. Matties, M.A. and Hentschke, R. (1996). Molecular dynamics simulation of benzene on graphite. 1. Phase behavior of an adsorbed monolayer. Langmuir, 12, 2495-500. 37. Matties, M.A. and Hentschke, R. (1996). Molecular dynamics simulation of benzene on graphite. 2. Phase behavior of adsorbed multilayers. Langmuir, 12, 2501-4. 38. Steele, W.A., Vernov, A., and Tildesley, DJ. (1987). Studies of the adsorption of N 2 on the graphite basal plane by computer simulation. Carbon, 25, 7-17. 39. Evans, DJ. and Morriss, G.P. (1983).Isothermal isobaric molecular dynamics ensemble. Chem. Phys., 77, 63-6. 40. Nose, S. (1984). A molecular dynamics method for simulations in the canonical ensemble. Mol. Phys., 52, 255-68. 41. Hoover, W. G. (1985). Canonical dynamics: equilibrium phase-space distributions. Phys. Rev. A, 31, 1695-7. 42. Nose, S. (1984). A unified formulation of the constant temperature molecular dynamics methods.]. Chem. Phys., 81, 511-19. 43. Kim, H.-Y. and Steele, W.A. (1992). Computer-simulation study of the phase diagram of the CH 4 monolayer on graphite: corrugation effects. Phys. Rev. B, 45, 6226-33. 44. Kofke, D.A. (1993). Gibbs-Duhem integrations: a new method for direct evaluation of phase coexistence by molecular simulations. Mol. Phys., 78, 1331-6. 45. Kofke, D.A. (1993). Direct evaluation of phase coexistence by molecular simulations via integration along the coexistence line.]. Chem. Phys., 98, 4149-62. 46. Widom, B. (1963).Some topics in the theory offluids.]. Chem. Phys., 39, 2802-12. 47. Finn, J .E. and Monson, P.A. (1988). Adsorption equilibria in an isobaric ensemble. Mol. Phys., 65, 1345-61. 48. Paredes, J.I., Suarez-Garcia, F., Villar-Rodil, S., et al. (2003). N 2 physisorption on carbon nanotubes: computer simulation and experimental results.]. Phys. Chem. B, 107,8905-16. 49. Martinez-Alonso, A. Tasc6n, J.M.D., and Bottani, EJ. (2000). Physisorption of simple gases on C 60 fullerene. Langmuir, 16, 1343-8. 50. Panagiotopoulos, A.Z. (1987). Direct determination ofphase coexistence properties of fluids by Monte Carlo simulation in a new ensemble. Mol. Phys., 61, 813-26. 51. Panagiotopoulos, A. Z. (1987). Adsorption and capillary condensation of fluids in cylindrical pores by Monte Carlo simulation in the Gibbs ensemble. Mol. Phys., 62,701-19. 52. G6zdz, W.T., Gubbins, K.E., and Panagiotopoulos, A.Z. (1995). Liquid-liquid phase transitions in pores. Mol. Phys., 84, 825-34. 53. Pierce, M. and Manousakis, E. (1998). Phase diagram of second layer of 4He adsorbed on graphite. Phys. Rev. Lett., 81, 156-9.
References
101
54. Pierce, M. and Manousakis, E. (1999). Path-integral Monte Carlo simulation of the second layer of 4He adsorbed on graphite. Phys. Rev. B, 59, 3802-14. 55. Pierce, M. and Manousakis, E. (1999). Monolayer solid 4He clusters on graphite. Phys. Rev. Lett., 83, 5314-17. 56. Nho, K. and Manousakis, E. (2002). Submonolayer molecular hydrogen on graphite: a path integral Monte Carlo study. Phys. Rev. B, 65, 115409-1-12. 57. Nho K., and Manousakis, E. (2003). Commensurate-incommensurate transitions in quantum films: submonolayer molecular hydrogen on graphite, Phys. Rev. B, 67, 195411-1-7. 58. Wang, Q. and Johnson, J.K. (1998). Hydrogen adsorption on graphite and in carbon slit pores from path integral simulations. Mol. Phys., 95, 299-309. 59. Wang, Q. and Johnson, J.K. (1999). Molecular simulation of hydrogen adsorption in single-walled carbon nanotubes and idealized carbon slit pores.]. Chern. Phys., 110, 577-86. 60. Wang, Q. andJohnson,J. K. (1999). Computer simulation of hydrogen adsorption on graphitic nanofibers.]. Phys. Chern. B, 103,277-81. 61. Challa, S.R., Sholl, D.S., and Johnson, J.K. (2002). Adsorption and separation of hydrogen isotopes in carbon nanotubes: multicomponent grand canonical Monte Carlo simulations.]. Chern. Phys., 116,814-24. 62. Gatica, S.M., Stan, G., Calbi, M.M., et al. (2000). Axial phase of quantum fluids in nanotubes.]. Low Temp. Phys., 120,337-59. 63. McGreevy, R.L. and Pusztai, L. (1988). Reverse Monte Carlo simulation: a new technique for the determination of disordered structures. Mol. Simul., 1, 359-67. 64. Pikunic, J., Clinard, C., Cohaut, N., et al. (2002). Reconstruction method for the characterization of porous carbons. Stud. Suif. Sci. Catal., 144, 19-26. 65. Pikunic, J., Clinard, C., Cohaut, N., et al. (2003). Structural modeling of porous carbons: constrained reverse Monte Carlo method. Langmuir, 19, 8565-82. 66. Thompson, K.T. and Gubbins, K.E. (2000). Modeling structural morphology of microporous carbons by reverse Monte Carlo. Langmuir, 16, 5761-73. 67. Pikunic, J., Pollen, J.-M., Thompson, K.T., et al. (2001). Improved molecular models for porous carbons. Stud. Suif. Sci. Catal., 132, 647-52.
MODELS OF POROUS CARBONS Henry Bock, Keith E. Gubbins, and Jorge Pikunic Department of Chemical and Biomolecular Engineering, North Carolina State University, Raleigh, NC, USA
Contents 5.1 Introduction 5.2 Experimental Probes
1°3 1°4
5.3 Molecular Models of Carbons 5.4 Adsorption, Diffusion, Reaction 5.5 Conclusions Acknowledgments References
106
5.1
121 127 128 128
INTRODUCTION
Except for the fullerenes, carbon nanotubes, nanohorns, and schwarzites, porous carbons are usually disordered materials, and cannot at present be completely characterized experimentally. Methods such as X-ray and neutron scattering and high-resolution transmission electron microscopy (HRTEM) give partial structural information, but are not yet able to provide a complete description of the atomic structure. Nevertheless, atomistic models of carbons are needed in order to interpret experimental characterization data (adsorption isotherms, heats of adsorption, etc.). They are also a necessary ingredient of any theory or molecular simulation for the prediction of the behavior of adsorbed phases within carbons - including diffusion, adsorption, heat effects, phase transitions, and chemical reactivity. Because the chemical and physical processes involved in the synthesis of disordered porous carbons are not well understood, attempts to develop mimetic modeling procedures, in which theory or simulation methods are used to mimic Adsorption by Carbons ISBN: 978-0-08-044464-2
© 2008 Elsevier Ltd. All rights reserved.
103
Chapter 5 Models of Porous Carbons
1°4
the complete synthesis, are difficult or impossible. There have been a few attempts to use ab initio or semiempirical methods to study some particular part of the synthesis route (see Section 5.3.3.2 below). Nevertheless, the great majority of attempts to model porous carbons can be classified as reconstruction methods, in which an atomistic model of the carbon is constructed that is consistent with available experimental structure data, usually X-ray scattering or transmission electron microscopy (TEM). The simplest reconstruction models are single-pore models, of which the slitpore model is the most widely used. The pore structure is represented as a collection of single, nonconnected pores of varying width, and in some cases varying wall thickness or wall heterogeneity. In the slit-pore model, the pores have parallel walls composed of graphene sheets; frequently these are treated as smooth and structureless. This model is widely used to estimate the pore size distribution, feR), where H is pore width, by assuming that the adsorption isotherm is a linear combination of isotherms calculated for pores of different sizes. While convenient, the slit-pore model neglects many important features of real disordered carbons, including pore connectivity, wall roughness, curved and defective graphene sheets, pores of wedge, and other shapes. These neglected features are known to play an important role in many processes in adsorbed phases, such as diffusion, phase changes, and chemical reactions. More recently, more sophisticated atomistic modeling methods have been proposed for disordered carbons that take these features into account at the cost of increased complexity and the need for more experimental data. In this chapter we give an account of the most useful and promising current methods for modeling porous carbons (Section 5.3). For completeness of the discussion we also include (Section 5.2), a brief account of the most important experimental methods for characterizing carbons. Following the section on models, we provide a review ofsome ofthe recent work on predicting the behavior of adsorbed phases within such model carbons (Section 5.4), including adsorption, heats of adsorption, diffusion, and chemical reactions. A more comprehensive review of the types of carbons, experimental methods for studying them, and molecular models has appeared [1] and covers the literature up to 2000.
5.2
EXPERIMENTAL PROBES
In 1917 Debye and Scherrer performed the first powder X-ray scattering experiments on diamond, graphite, and several amorphous carbons (chars) [2]. From their experimental results, Debye and Scherrer concluded that amorphous carbons consist of small graphitic units of about 30 atoms. Scattering is sensitive to the distance between the scattering particles. If a certain distance, d, appears in a periodic fashion, we find constructive interference of the scattered radiation of wavelength, A, at an angle, 2fJ with respect to the incident radiation following Bragg's law, d = A/2sin fJ. The scattering pattern of carbon blacks and activated
5.2 Experimental Probes
105
carbons consist of the (002) three-dimensional (3D) peak and several (hk) twodimensional (2D) peaks. The (002) 3D peak indicates stacking of individual graphitic layers and allows the determination of the inter layer distance, d = 0.32-0.38 nm; 0.3354 nm is found for graphite. The absence offurther 3D peaks is caused by horizontal distortion of the individual graphene sheets within the stack, i.e., by turbostratic ordering. The 2D peaks provide information about the ordering within the graphene sheets and the carbon-carbon distance (0.1415 nm for graphite). Porous carbons generally show less inlayer ordering than graphite and usually a larger interlayer spacing, d002 . The dimensions of the graphene stacks can also be determined from the scattering data. For activated carbons, a stack height of 1 nm (2-3 layers) and a stack width of 1-3 nm is found. Strong scattering but no peaks in small-angle X-ray scattering (SAXS) (2fJ < 15°) indicate the existence of a nonperiodic pore network at length scales beyond the interlayer spacing [3]. Information about pore morphology and topology, i.e., about the relative position and orientation of the graphitic crystallites and their connectivity is accessible using TEM or HRTEM. The basic principles of TEM are similar to those of conventional light optical microscopy. The contrast in TEM originates from scattering of electrons on the atoms of the porous carbons. Scattered electrons do not pass through the small opening of the objective aperture, thus denser regions of the carbon appear darker in the TEM image since they cause more scattering. In TEM, only the direct beam passes trough the aperture; while in HRTEM also diffracted beams contribute to imaging. This results in a higher resolution and allows not only the determination of qualitative information such as shape and orientation of the pores but also quantitative information such as pore size distributions. It is even possible to obtain the interlayer distance, d002 . Since the graphene sheets in activated carbons are finite and have defects, as seen from X-ray scattering, it is obvious that they are terminated by heteroatoms, such as H, 0, N, S, and P. These atoms generate heterogeneities in the walladsorbate interaction and thus their inclusion is essential for the adsorption properties of carbon materials. To include heteroatoms in models of amorphous carbons, we must know which and how many functional groups a specific carbon material contains and where they are located. Table 5.1 lists a number of methods which can be used to identify these functional groups. The Boehm titration is a wet chemical method where a sequence of bases with increasing basicity (pKa ) is used to titrate (neutralize) substances having a pKa less or equal to that of the base. Thus it is a reliable method to obtain general information about the surface acidity. Modern potentiometric titration significantly improves the pKa resolution compared to the Boehm titration. The usage of nonaqueous solvents as in calorimetric titration increases the pKa range that can be probed. All titration methods are, however, insensitive to the chemical nature of the functional groups, only their acidity is measured. X-ray photoelectron spectroscopy (XPS) uses X-rays to eject core electrons (ls) from carbon, oxygen, nitrogen, or other heteroatoms. The binding energy
Chapter 5 Models of Porous Carbons
106
Table 5.1 Methods for the characterization of activated carbons (Adapted from Ref. [1].)
Small-angle X-ray scattering X-ray diffraction Transmission electron microscopy Gas adsorption Boehm titration Potentiometric titration Calorimetric titration Temperature-programmed desorption (TPD) Fourier transform infrared spectroscopy X-ray photoelectron spectroscopy Immersion calorimetry Flow adsorption Inverse gas chromatography
Total surface area (open and closed pores) mean pore size Mean crystallite size, carbon-carbon pair correlation function 2D images of the pore structure, matrix correlation function Porosity, pore size distributiona , surface area a pKa of functional groups, their number pKa of functional groups, their number pKa of functional groups, their number Type of oxygen functional groups (strong or weak acids)/their number Type of functional groups (number) Type of functional groups, amount of heteroatoms (number of groups) Number of primary adsorption centers (oxygenated groups) Average polarity Average acidity
aThe introduction of an approximate model or a theory is needed to extract this data from the experiments.
of the core electron is measured. Thus XPS is sensitive to the atom type and to the way this atom is bound to its environment. Fourier transform infrared spectroscopy (FTIR) probes molecular vibrations. FTIR spectra are usually analyzed qualitatively by comparison with FTIR spectra of known organic compounds. Gas adsorption is sensitive to various properties of carbons, such as pore size, surface area, and porosity. Thus, it can be used itself to analyze carbon materials. However, the data interpretation relies on appropriate models to connect the adsorption results to properties of the carbon material (see Section 5.3).
5-3
MOLECULAR MODELS OF CARBONS
5.3.1 Regular Porous Carbons Regular porous carbons are carbon materials with a simple pore geometry; they include carbon nanotubes, fullerenes, and schwarzites. If carbon nanotubes are considered for sensor applications, ab initio models are necessary to test whether adsorption of a certain molecule (e.g., N0 2 ) generates a change of the electron density of the nanotube that is large enough to use the nanotube in a sensor [4].
5.3 Molecular Models of Carbons
107
Classical models of carbon nanotubes can be subdivided into two groups: explicit geometric and mean-field models [1]. An explicit model for a single-wall nanotube is obtained by rolling-up a graphene sheet (see Section. 5.3.2) to form a cylindrical surface. Geometric constraints determine possible nanotube diameters, unit cells, and symmetries (zigzag or armchair) [5, 6]. For the carbon atom-adsorbate atom potential [6] the Lennard-Jones (LJ 12-6) potential is often used. If the adsorbate is not sensitive to the atomistic details of the nanotube, e.g., the adsorbate atoms are much bigger than the carbon atoms, a mean-field model can be derived [7]. The mean-field potential is obtained by representing the nanotube wall by an areal density of carbon atoms (LJ 12-6) rather than their explicit positions and integrating along the azimuthal and longitudinal direction (see also Section 5.3.2). The resulting potential depends only on the (normal) distance between the adsorbate and the nanotube. The integration can also be performed numerically, which might be advantageous if more complicated nanotube structures are considered [8]. The extension to multiwall nanotubes and bundles or ropes is straightforward. One obtains the total adsorbate-nanotube potential by superposition of the individual single-wall nanotube potentials. Especially in heterogeneous bundles it is necessary to find the equilibrium configuration of the bundle [9]. In the case of explicit models the relative orientation (rotation) of the individual walls of multiwalled nanotubes and the relative orientation (rotation) of individual nanotubes in a bundle must be decided too [1]. Another interesting polymorph of carbon is fullerene. Although adsorption on individual fullerene molecules and on the surfaces offullerene crystals is not widely studied, explicit [10] as well as mean-field [11] models are available for individual fullerene molecules. The fullerene crystal can be modeled by placing individual (model) fullerene molecules on the sites of an fcc lattice, to match the symmetry and density ofthe real solid [12] or to match equilibrium structures obtained from computer simulations offullerene crystals [10]. A model for a defective crystal can be obtained by removing some ofthe fullerene molecules [13]. Besides fullerenes, nanotubes, and graphite, which are finite or (quasi) infinite in one or two dimensions, regular carbon materials that are infinite in three dimensions, called schwarzites, exist. Schwarzites can be synthesized inside zeolites or other ordered porous silica materials [14]. Thus, their topology and morphology is similar to that of the template, i.e., they consist of an extended network of channels and cages, each one being separated from a neighboring one by a graphene-like wall comprising five- and eight-membered rings. Explicit models for several schwarzites are available in the literature [14].
5.3.2 Disordered Porous Carbons: Simple Geometric Models The evolution of molecular models for disordered porous carbons is strongly connected with the advance of experimental techniques such as diffraction methods and electron microscopy. First, X-ray studies on carbon blacks revealed that these materials consist of a wealth of small graphitic crystallites.
108
Chapter 5 Models of Porous Carbons
The spatial arrangement of these graphitic crystallites determines the pore structure, i.e., the pore morphology and topology. As a consequence, all simple models of porous carbons are based on stacks of graphene sheets representing the pore walls. As in the case of nanotubes discrete as well as mean-field models of single graphene sheets or stacks of them are available. In several cases, both descriptions have been combined in hybrid models. A discrete model of a graphene sheet is obtained by placing (model) carbon atoms on a - 2D hexagonal lattice with lattice constant (carbon-carbon bond length) e. The values of e measured in disordered carbons are usually very similar to that of graphite. Thus, the carboncarbon bond length of graphite, e = 0.1415 nm, is usually used to construct the simple models discussed here. Carbon atoms as well as the adsorbate molecules are often modeled by the LJ 12-6 potential, (5.1) where Gfc and O"fc are the fluid-carbon energy and distance parameters, l respectively, and r is the distance separating two atoms or molecules. The total fluidwall interaction energy of an adsorbate molecule, j, with the graphene sheet is given by the sum, Li cP L] (rij) , which runs over all wall (carbon) atoms, i. This readily defines an explicit model for adsorption on a single graphene sheet. If we can disregard the atomistic nature of the graphene sheet, the sheet is sufficiently characterized by an areal carbon density. A mean-field model is obtained by replacing the sum over individual carbon atoms by an integral of the LJ potential over the area of the graphene sheet. If the sheet is assumed to be infinite in lateral dimensions one obtains the well-known Steele (10-4) potential [15] (5.2) where Psd is the areal number density of carbon atoms in the graphene layer. (Since the structure of graphite is known, the number density of carbon atoms in graphite can be calculated: Ps =2.0/(d12 J'3(3/2)) ~ 114nm- 3 .) With similar integration methods one obtains mean-field potentials for graphene sheets that are finite in one or more directions [1]. A model for a finite stack of graphene layers is obtained by summing over individual graphene layers modeled as described above. The spacing between the layers, d, is usually assumed to be that of graphite, d = d002 = 0.3354 nm. 1
Parameters for LJ potentials between unlike particles are usually obtained from the LJ potentials for the like-like interactions using Lorentz-Berthelot mixing rules, Ore = ,J0ffOee and arc = 1/2 (aff + aeJ. For graphite 0ee ~ 38.7 x 10- 23J and ace ~ 0.34 nm. A list of adsorbate parameters can be found in Ref [15].
5.3 Molecular Models of Carbons
1°9
For simple models this is reasonable, since deviations from this value measured for activated carbons are usually smaller than 0.03 nm and structural simplifications may be more serious. In explicit models ofgraphene stacks the registry ofadjacent layers has to be considered. The thermodynamically stable polymorph ofgraphite is hexagonal graphite having an ABA stacking. The position of the B layer with respect to an A layer can be obtained by starting from a perfectly aligned system (AAA) and displacing every second layer along an arbitrary bond by one bond length. In hybrid models the structure of the first few layers is considered explicitly, while for all other layers a mean-field description is adopted. To derive an even simpler description of a semi-infinite graphite substrate that is infinite in all directions but semi-infinite in the direction perpendicular to the graphene layers, the substrate is represented by a 3D carbon density. Volume integration gives the Steele (9-3) potential [15],
(5.3) Another mean-field potential, which is often used, is obtained by areal integration over the first layer of a semi-infinite graphite substrate and volume integration over all others. This gives the well-known (10-4-3) Steele potential [15],
U z (
) -
211'8
fcPc
[2
10
4
4]
afc a 2 d - -afc - -afc fc 5( z ) ( z) 3d(z+O.61d)3
(5.4)
These different wall models can now be assembled such that they confine some spatial region which represents the pore. The simplest and most widely used case is the slit pore. A slit pore of width, H, is created by placing two mutually parallel and laterally infinite graphene stacks at a distance, H, apart. Pores shaped like rectangular prisms are obtained by placing four graphene stacks at the side faces of the prism. To avoid overlap the graphene stacks are laterally finite or semi-infinite in the dimension perpendicular to the long axis of the prism [16, 17]. TEM images of many specimens reveal that the pore walls are usually not parallel. Because of this observation, a model for pores shaped like triangular prisms has been developed [18]. It is important to notice that the latter two models differ from the slit-pore model not only by pore morphology but also by the appearance of high-energy sites at the edges, generated by proximity of the two walls. Hybrid models are usually used to study defective surfaces. In these models the internal surfaces of a slit pore, defined by a stack of mean-field layers, are "coated" by one or more graphene layers with explicit atomic structure. To generate defects one or more carbon atoms are removed from one or more explicit graphene layers [19, 20]. All models discussed in this section are single-pore models describing the morphology of a single pore. The topology of the disordered material,
110
Chapter 5 Models of Porous Carbons
i.e., connectivity of pores, as well as the variety of pore sizes is completely disregarded, which significantly limits the predictive capabilities of these models. In recent years experimental techniques have been improved, providing much more detailed information of carbon materials. As described in the following section, this information is used together with sophisticated theoretical and simulation methods to obtain more detailed and more reliable models for disordered porous carbons.
5.3.3 Disordered Carbons: More Realistic Models 5.3.3.1 Reconstruction methods The goal of reconstruction methods is to build model pore structures that match experimental structure data (including surface chemistry data) for the real materials, at least in a qualitative way. For example, models can be constructed that match the experimental structure factor, Seq), or TEM data, by reverse Monte Carlo (RMC), off-lattice reconstruction, or other methods. Reconstruction methods can in principle be used to build a model for any type of porous material. However, considerable care and thought is needed in applying such methods, since the experimental data does not correspond to a unique molecular structure. For example, a range of atomic structures could give rise to the same S( q) curve or TEM data. This ambiguity can be reduced by incorporating constraints into the model development, so that unphysical structures cannot result, and by using more than one kind of experimental data in the fitting process. The simple geometric models mentioned above (see Section 5.3.2) can be thought of as the most basic form of reconstruction methods, in which the pore topology is based on electron micrographs of the material (e.g., cylindrical, slit-shaped), and the structure of the pore walls is assumed to be that of graphite. In the case of activated carbons, it is common to use a slit-pore model with graphitic pore walls. The only adjustable parameter is then the pore size, which may be estimated from electron micrographs. This model may be refined by including a distribution of pore sizes. The porous material is then modeled as a collection of independent and unconnected slit-shaped pores with graphitic pore walls, and a pore size distribution is determined so that the calculated adsorption isotherm matches the experiment [21]. There have been several improvements to the slit-pore model and the description based on the concept of a pore size distribution. These improved models are also constructed by making detailed observations of the experimental data (electron micrographs, X-ray diffraction, adsorption isotherms, etc.), extracting more relevant features of the pore topology and the structure of the pore walls, and including these features in the models. For example, a 2D distribution of pore size and pore-wall thickness may be used, instead of a pore size distribution [22]. Most of these improvements are described in detail in a recent review [1].
5.3 Molecular Models of Carbons
111
The recent trend, driven by the increase in computing power, is to build all-atom models of porous carbons by solving a multidimensional inverse problem. The atomic positions in a system of carbon atoms, and perhaps other species, are stochastically changed to match experimental structure data. The RMC procedure [23] is useful to produce configurations that match a target structure factor, S(q), or pair correlation function, g(r). Target structure factors are usually obtained from X-ray diffraction. In addition, SAXS may be used to extend the structure factor to lower values of q, provided that X-ray diffraction and SAXS can be performed for an overlapping range of angles [24]. Target pair correlation functions may be obtained by taking the inverse Fourier transform of the structure factor. This operation, however, is particularly vulnerable to the limitations of the experimental data [25, 26], e.g., truncation errors. It is thus preferable [27] to use an alternative method, such as the so-called MC g(r) [26]. The idea of this inverse procedure is to stochastically modify a pair correlation function until its Fourier transform matches the experimental structure factor. The procedure is analogous to a lD RMC. Since the numerical pair correlation function can be generated for arbitrarily large r-values (limited only by the experimental q resolution), truncation errors are avoided in the Fourier transform. The resulting pair correlation function may be used as the target function in the RMC procedure, instead of the experimental structure factor, significantly reducing the computational cost and thus allowing the study of larger systems. In RMC, random moves, i.e., changes in the configuration of the system, are performed as in the metropolis MC algorithm. Random moves are accepted or rejected so that the difference between the calculated and target S( q) or g(r) is minimized. If g(r) is used as the target function, the parameter to be minimized is: nexp
L X2 =
[gsim (rJ -
gexp
(rJ]2
i=l
- - n -- - - - - - ex p
L
(5.5)
[gexp (rJt
i=l
where nexp is the number of experimental points, gsim (ri ) is the simulated g(r) and gexp (ri) is the experimental g(r) evaluated at rio The moves are accepted with probability:
Pace
= min [ 1, exp { -
;x (X~ew - X~ld) }]
(5.6)
where the subscripts old and new indicate before and after the move, respectively, and Tx is a weighting parameter or effective temperature. It is important to note that, although the parameter Tx does not have a thermodynamic meaning, it behaves like temperature. The parameter X is therefore minimized when the effective temperature is close to zero. The original RMC procedure prescribes that T x should be proportional to the variance of the target function.
112
Chapter 5 Models of Porous Carbons
However, assuming that the error in the target function is relatively low, any arbitrarily low value would be appropriate. An alternative is to change the effective temperature in the frame of the simulated annealing method [28] to increase the chances of finding the global minimum of the parameter x, instead of a local minimum. Simulated annealing has been successfully used in RMC modeling of carbons [27, 29]. As mentioned above, a set of experimental data does not necessarily correspond to a unique molecular structure. Moreover, even unphysical structures may be consistent with a set of experimental data. It is therefore necessary to carefully choose a set of constraints to limit the number of possible structures. The uniqueness theorem of statistical mechanics [30, 31] provides a guide to the number and type of constraints that should be applied in the RMC method in order to get a unique structure [32]. For systems in which only two- and three-body forces are important, the uniqueness theorem states that a given set of pair correlation function and three-body correlation function determines all the higher correlation functions. In other words, assuming that only two- and three-body forces are important, the RMC method must be implemented along with constraints that describe the three-body correlations [27]. One way ofimplementing constraints is in a rigid way. For example, Thomson and Gubbins [33] have modeled an activated mesocarbon microbead with RMC along with the following constraints: (1) any atom can only have two or three neighbors, (2) all the carbon-carbon bond lengths are 1.42 A, and (3) all the bond angles are 120 The advantage of doing this is that when these three constraints are applied together, basic carbon units, or plates, can be defined. These plates are rigid aromatic sheets of Sp2-bonded carbon, which resemble the structure of graphene segments. Many-atom moves that accelerate the convergence process can then be applied. For example, Thomson and Gubbins included three types of stochastic moves: (1) plate translation-rotation, (2) ring creation-annihilation, and (3) plate creation-annihilation. Only those moves that improve the fit to the experimental radial distribution function are accepted, i.e., the effective temperature was set to zero. In their resulting models, the graphene segments are roughly aligned (see Fig. 5.1) but their shape, size, and relative angles of tilt are different. The match between the simulated and the experimental g(r) is excellent for interatomic distances greater than 5 A (see Fig. 5.1). However, deviations occur at smaller distances. Two possible reasons for this discrepancy are (1) truncation errors in the Fourier transform of the experimental structure factor to obtain the radial distribution function, producing unphysical features in the target radial distribution function, and (2) overly rigid constraints on the RMC platelet shape. While the constraints applied by Thomson and Gubbins are reasonable for many graphitizable carbons, the use of graphene microcrystals as the basic units fails to account for ring defects and nonaromatic rings that are important in many activated carbons used in adsorption applications. A better match to the target g(r) may be obtained by allowing the formation of defects in the form of nonaromatic rings and by including heteroatoms [34]. 0
•
5.3 Molecular Models of Carbons
113
(a)
(b) 10
5 ~
(J) c:
(J)
~
0
-------t
+--~--t-----t--&------t==......
C (J) (5
a..
a _.....-+-_--1.--
0.0
Evib
aEovib
0.5 1.0 1.5 2.0 2.5 3.0 3.5 Distance from surface (molecular diameters)
4.0
Figure 7.2 A schematic diagram of the various energy changes taking place on adsorption.
15 1
7.2 Theoretical Background
more suitable than any of the experimentally measured heats as an index of the fundamental "affinity" of a solid surface for adsorbing a particular gas molecule. The relation of the various heats of adsorption to the adsorptive potential is shown schematically in Fig. 7.2. It should be noted that while heat lost from a system is thermodynamically a negative quantity, it is a custom of long-standing to employ a positive sign in adsorption science. This is frequently confusing to newcomers to adsorption studies. For a more complete and detailed discussion of the thermodynamic quantities ofinterest in physical adsorption, the reader is referred to Chapter III of Ref [3].
7.2
THEORETICAL BACKGROUND
7.2.1 The Integral Equation of Adsorption
Although the concepts are somewhat older, the most widely used model for describing adsorption on an energetically heterogeneous surface was first explicitly stated by Ross and Olivier [4, 5]. The model postulates that the surface' of a real solid is composed of small patches of different adsorptive potential that adsorb independently of one other. The distribution of adsorptive potentials, Uo, among these patches may be represented by a continuous distribution function:
1 da fa= AdU. =f(Uo)
(7.7)
°
where fa is the patch (or site) frequency per unit energy interval on a surface of area A. The distribution function must normalize to unity, as was pointed out by Hill [6], since we are dealing with a surface of finite extent; that is, !(Uo)dUo = 1, over the range of energies considered significant. At any equilibrium pressure p under isothermal conditions, the quantity adsorbed per unit area, q, on a given surface patch will depend only on the adsorptive potential of that patch according to a function: or more generally q = q (P, Uo)
(7.8)
The observed total amount adsorbed, Q at pressure p is then the sum of the contributions from each patch of surface, i.e., (7.9) Equation (7.9) is therefore the general form for any adsorption isotherm and corresponds to equation IV-4 ofRef [3]. Equation (7.9) is now often referred to as "the integral equation of adsorption" or "the generalized adsorption integral." The function q(p,Uo) is called the kernel function or the local isotherm. The local isotherm can take various forms, depending on the geometry of the system that Eqn (7.9) is being used to describe.
Chapter 7 The Surface Heterogeneity of Carbon and Its Assessment
7.2. 2 Solving and Using the Integral Equation of Adsorption 7.2.2.1 Analytic solutions
Referring to Eqn (7.9), we see that in any treatment of surface heterogeneity, we have to deal with three functions, any two of which, if known, assumed or determined can be used in theory to obtain the third. Equation (7.9) represents a Fredholm's integral of the first kind. The solution of equations of this type is well known to present an ill-posed or ill-conditioned problem. For our purposes, this means that the data, Q(P), can be "well represented" by many function pairs in the integrand; hence, simply fitting the data does not guarantee that the kernel function or the distribution are individually "correct." In addition, the mathematical difficulties of handling Eqn (7.9) analytically have severely restricted the number of possible variations that have been published and these are now only of historical interest. No analytic solution of Eqn (7.9) has yet been made based on reasonable models of multilayer adsorption incorporating adsorbate-adsorbate interaction; such a solution may not be possible. 7.2.2.2 Numerical solutions
Unlocking the utility ofEqn (7.9) has been a challenge for decades. The period of renewed adsorption research activity in the decade of the 1950s happened at a time when high-speed electronic computing was just becoming available to researchers in this field. This made the numerical solution of Eqn (7.9) a feasible undertaking. For the first time, it was possible to at least calculate the numerical values of Q(P) from the integral equation of adsorption using more theoretically sophisticated kernel functions that incorporated adsorbate-adsorbate interaction, together with a reasonable distribution function. Equation (7.9) can be rewritten in discrete form as a summation over all significant adsorptive potential patches: (7.10) where we have replaced Ua with the less specific equivalent, B i . The first such solutions were carried out by Ross and Olivier [4, 5] and are tabulated in Ref [3]. Using Gaussian distributions of adsorptive potential of varying width, they computed tables of model isotherms using kernel functions based on the Hill-de Boer equation for a mobile, nonideal two-dimensional gas. It was not actually possible to fit data to the computed models using numerical methods in 1957, so Ross and Olivier developed a technique to find the best fitting model for an experimental isotherm data set by means of graphical overlays. They found that excellent fits to the experimental data could be obtained provided that the degree of heterogeneity was not too great. As pointed out above, a good fit to the data does not in itself verify a kernel
7.3 The Application of Density Functional Theory
153
function or the distribution. However, as the adsorbent becomes more and more homotattic, its isotherms should approach agreement with the kernel function. This was shown to be the case for adsorption measurements on a series of carbons graphitized at increasing temperatures, culminating in the highly graphitized carbon black, P-33, whose isotherms of argon and nitrogen at 77 K and 90 K are closely fitted by the Hill-de Boer equation in the monolayer region. By also correctly describing the heat of adsorption as a function of quantity adsorbed for heterogeneous surfaces, this work confirmed Eqn (7.9) as a powerful tool for investigating surface heterogeneity and the validity of the two-dimensional nonideal gas model for the kernel function. In later work, Ross and Morrison [7, 8] were able to make several advances. The van der Waals equation of state for real gases, which is the basis of the Hill-de Boer equation, is known to be rather inaccurate. Ross and Morrison based their kernel function on a two-dimensional form of the much better virial equation of state. But more importantly, advances in computing resources made it possible to solve Eqn (7.10) for the unknown distribution function using a nonnegative least squares method, rather than assuming a form a priori [9]. Again, it was found to be difficult to fit uniquely isotherm data for surfaces that were more than moderately heterogeneous. The major limitation lies in the fact that the kernel functions used were only models for monolayer adsorption, yet it is well known that adsorption proceeds to multilayers as pressure is increased. To ensure that the more strongly adsorbing portions of the surface remained in the monolayer range, only the lowest pressure portion of the isotherm can be used. This means that the low adsorptive energy portions of the surface contribute little to the total amount adsorbed, making their estimation uncertain. If higher pressure data are included in Q(P), then multilayers exist on some surface patches, which are then not correctly modeled by the monolayer kernel function. Further advances had to await the theoretical development of an improved kernel function.
7.3
THE ApPLICATION OF DENSITY FUNCTIONAL THEORY
While good descriptions of adsorption on uniform surfaces in the submonolayer region have been available for decades, only since the 1990s has accurate calculation ofthe whole isotherm, including the multilayer region, been demonstrated [10]. These calculations use a modified nonlocal density functional theory (MNLDFT). The first use of multilayer local isotherms calculated by MNLDFT in obtaining a measure of surface energetic heterogeneity for several solid adsorbents was reported in 1996 [11]. The formalism of density functional theory (DFT) has received considerable attention as a way to describe the adsorption process at the fluid-solid interface. The older approach was to treat the adsorbate as a separate, two-dimensional phase existing in equilibrium with the bulk gas phase. This model works well
154
Chapter 7 The Surface Heterogeneity of Carbon and Its Assessment
in the monolayer region, but at higher surface concentrations the formation of multilayers requires adopting some sort of three-dimensional model in order to account for increasing adsorbate-adsorbate interaction and the diminishing adsorption potential contribution. Using density functional theory, the adsorptive can be treated as a single, inhomogeneous fluid phase. The fluid varies in density from that of the bulk gas to a much higher value at the adsorbent surface in response to the strength and configuration of the surface adsorptive forces. In this paradigm, there is no separate adsorbed phase; indeed, the concept of a monolayer capacity, fundamental to the two-phase paradigm has disappeared as well. The benefit of this approach is that the isotherm can be modeled from the Henry's law region through to saturation, and even above the adsorptive's critical temperature. In particular, the ability with DFT to model physical adsorption in a pore space of slit-like or cylindrical geometry has led to potentially useful methods for extracting surface area and pore size distribution information from experimental adsorption isotherms [12, 13]. The predictions of density functional theory have been reported to compare well with the results of simulations [14, 15] using Monte Carlo or molecular dynamics methods. Stringent comparisons to real data have been made by us [10] for the adsorption of nitrogen and argon on the near-homotattic surface of a highly gr~phitized carbon, Sterling FT-G(2700). In performing such comparisons, the only unknown intensive parameter is the LJ pairwise interaction energy X f between the adsorbate and adsorbent atoms. Using the customary Tarazona [16] prescription (with corrected weight functions [17]) for the free-energy density functional, we have found that the experimental isotherm data in the monolayer region of coverage can be moderately well described by DFT calculations; however, in the multilayer region of the isotherm, the quantity adsorbed per unit area is significantly over predicted. Later work [10] has shown that a modification to the mean field approximation used to calculate the attractive component of the configurational chemical potential leads to theoretical isotherms that agree closely with experiment over a six-decade range of pressure. An example is shown in Fig. 7.3, along with the results of the unmodified NLDFT of Tarazona [17].
7.3.1 The Deconvolution Method The integral equation of adsorption, Eqn. (7.9), can be rewritten in specific units as
Q(P)
=
f deq(p, e)j(e)
(7.11)
where Q(P) is the total quantity of adsorbate per gram of adsorbent at pressure p, q(p,X) , the kernel function (the local isotherm), describes the adsorption
isotherm for an ideally homotattic surface characterized by an interaction energy e as quantity of adsorbate per square meter of surface, and f(e) the surface area distribution function with respect to e. The quantity e(Eqn. 7.1)) as we
7.3 The Application of Density Functional Theory
0.5
~
g:
-r-----------------~_:r_'1
-0-
0.4
(J)
155
Data
--- MNLDFT --- NLDFT
('I)
E
~
0.3
-0
Q)
-eo ~
0.2
ECO
0.1
CO ~ :::J
o
0.0 ..................~~-........_--,.__--r__--.....__-___4 1e-6 1e-5 1e-4 1e-3 1e-2 1e-1 1e+0 Relative pressure
Figure 7-3 A comparison of experimental data for nitrogen adsorbed at 77 K on Sterling FT-G(2700) graphite with the modified nonlocal density functional theory [10] (MNLDFT) and unmodified [1 7] nonlocal density functional theory (NLDFT).
shall see is closely related to the adsorptive potential and can be equated to the quantity a pads in Eqn. (7.3) and in Fig. 7.2. While DFT allows us to calculate values for q(p, 8), it of course provides no analytic form for the function, and in general the form of f(8) is also unknown. However, by using carefully designed numerical methods, model isotherms calculated by MNLDFT can be used in carrying out the inversion of the discrete form of the integral equation of adsorption. In this way one can determine the effective adsorptive potential distribution of the adsorbent from the experimental adsorption isotherm. The method used can be expressed by (7.12) where Q(P) is the experimental adsorption isotherm interpolated onto the vector p of pressure points, q(p, 8 ij ) a matrix of quantity adsorbed per square meter, each row calculated by MNLDFT for a value of 8 at pressures p, andf(8J a vector of positive or null values whose terms represent the area of surface in the sample characterized by energy Xi' The total surface area of the sample is given by
The solution values desired are those positive numbers that most nearly, in a least squares sense, solve Eqn. (7.12). Additional constraints on the solution may be required to stabilize the deconvolution process [18, 19]. The formulation and solution of Eqn. (7.12) differs from previous work in an important way. In previous attempts, Q(P) was the amount adsorbed at the
Chapter 7 The Surface Heterogeneity of Carbon and Its Assessment
experimental pressures, p. This required that q(p, XJ be calculated for that specific set of pressures, and that the size ofJ (XJ be no greater than the number of experimental points. Not only does this result in a large computing task for each data set, it causes the evaluation ofJ(XJ to be subject to a varying bias, depending on how many and where on the pressure scale the isotherm points were measured. The automatic adsorption equipment available today permits a large number of experimental points to be measured and the resulting isotherms can be interpolated accurately onto a predetermined optimized set of pressures. Hence, the vector p can be chosen to best represent the kernel function over the wide pressure range required by the set of Xi. If we consider m members of the set of X and a vector p of length n, it is clear that n X m must hold. If n = m, the solution vector J (X) can be "noisy" because of even small imperfections in the data or model. For n > m, the solution is smoother because of the additional data constraints. Various other regularization techniques have been proposed to mitigate the inversion problem; in this work we use the method of co-minimization of the second derivative of f(X) together with an overdetermined matrix for which n > 2m.
7.3.1.1 Parameters of the model matrix For use with Eqn.(7.12), a model matrix was calculated by the MNLDFT method [10] using the parameters suitable for nitrogen at 77.3 K. For convenience, the values of X were specifically the values of X sf / k used to calculate the wall potential V(z) of that reference, and ranged from 20 to 100 K in steps of 2.0 K (approximately 1/2RT). Relative pressure points were chosen in geometric progression from 1 x 10-6 to 0.6 with 40 points per pressure decade. Model isotherms were normalized to 1 m 2 of surface.
7.4
RESULTS FOR "NONPOROUS" CARBONS
Synthetic and natural graphites and carbon blacks are arguably nonporous, though the small spaces between primary particles in a carbon black agglomerate may act as pores in some materials. Additionally, the prismatic surfaces of natural graphite may display "missing" graphene planes that in effect become shallow slit-like pores. If such pores have a width less than about 1 nm, they will report as very high energy regions in the adsorptive energy distribution. The data reported here were obtained using a Micromeritics ASAP 2010 equipped with optional 10- and 1-torr pressure transducer. Low-pressure data were corrected for thermal effusion.
157
7.4 Results for "Nonporous" Carbons
7.4.1 Synthetic Graphitic Carbons Heating a graphitizable thermal carbon black to high temperature in an inert atmosphere produces some of the most energetically uniform surfaces known. One reason for this lies in the shape of the particles formed. Electron micrographs [3] reveal that the individual particles are doubly truncated polygonal (principally hexagonal and octagonal) bipyramids consisting of minute radiating crystals. The exposed surface of each crystal is the graphite basal plane. The surface of the whole faceted particle is therefore entirely composed of the carbon layer plane of crystalline graphite with no exposed prismatic surface. To confirm the deconvolution algorithm, we show in Figs. 7.4(a) and (b) the result of applying Eqn. (7.12) to the experimental data contained in Fig. 7.3. Since this data set was used in developing the MNLDFT model, we would expect to recover a monomodal energy distribution with esf / k = 57.0 K, as used in the fit shown in Fig. 7.3. The best fit contained contributions from the classes representing esf / k = 56 and 58, with an area weighted mean of 56.7 K, which is satisfactory agreement. The total surface area obtained is 12.4m2 jg. The BET (stands for Brunauer, Emmett, and Teller) area of this certified reference material is 11.1 m 2 j g. Because the MDFT model ignores the slight corrugation of the wall potential, the commensurate film transition seen at 0.008 relative pressure is not reproduced. Figure 7.5 , (a) and (b), illustrates the application of Eqn. (7.12) to the nitrogen isotherm obtained with Vulcan 3-G(2700). While graphitized at the same temperature as the Sterling FT, Vulcan has previously been reported as less uniform than the Sterling material [14]. As additional evidence, note that the commensurate film transition near 0.008 Pre! seen in Figs. 7.3 and 7.4 is not
(a) 6 - r - - - - - - - - - - - - - - - - - , -
f
(b) 10 . , . . . . . - - - - - - - - - - - - - - - ,
5 •
~
en 4
-
Experimental data MNLDFT fitted
8
ME
()
:;; 3 Q)
.c
~
"'C
2
eel
~1 c: eel
2
::J
o
0
1e-6 1e-5 1e-4 1e-3 1e-2 1e-1 1e+0 Relative pressure
0+--.,.----r------..LL1r----r---r--------l o 20 40 60 80 100 120 csf/k (K)
Figure 7.4 (a) The expe9mental data (points) of Fig. 7.3 fitted by Eqn (7.12) using the deconvolution method (solid line). (b) The adsorptive potential distribution for the Sterling graphite.
Chapter 7 The Surface Heterogeneity of Carbon and Its Assessment
(a) 40 - r - - - - - - - - - - - - - - - - , 1 E " " " 1 (b) 40 - , - - - - - - - - - - - - - - - ,
~ 0..
~ (f)
•
35 -
Experimental data DFTfitted
~ S
30
,£.
Q)
u
"0
~ 20
~ 20
o
::J CIJ
CIJ
~
«;
15
'E Q)
~
~
E ~
10
::J
o
30
co co ~
ME 25
10
(.)
E
5
o -+-------r::::y;.;...=..---r-----r-----r----.----f 1e-6
1e-5
1e-4
1e-3
1e-2
1e-1
1e+0
O-t----.,.---20 40
80
100
Relative pressure
Figure 7.5 (a) A comparison of experimental data for nitrogen adsorbed at 77 K on Vulcan 3-G(2700) (points) with the fit given by the modified nonlocal density functional theory (MNLDFT) models (line). (b) The adsorptive potential distribution for the Vulcan 3 graphite.
experimentally detected on the Vulcan surface. The area weighted mean value ofXsf / k is 56.1 K, and the total reported surface area is 80.78 m 2 I g. The surface area by BET is 73.5 m 2 /g.
7.4.2 Natural Graphites Natural graphites differ from those described above chiefly in their morphology. While equally crystalline, virtually all possible growth, cleavage, and fracture surfaces are present along with the extended basal surfaces. Interest in characterizing these materials has grown because of their importance in batteries for light-weight energy storage. The performance of a graphite anode in a lithium ion battery is known to be strongly related to the graphite's surface properties, in particular to the surface area and to the relative extent of basal plane and prismatic crystallite surfaces exposed to the electrolyte [20]. The presence of prismatic surface is necessary to allow the intercalation of the Li+ ion into the bulk of the graphite. In principle, therefore, graphites with a higher ratio of prismatic to basal surfaces should yield superior performance. We illustrate this in Fig. 7.6, (a) and (b). We see that the fine-grinding procedure has the expected result of increasing the total surface area, from 6.28 to 25.78 m 2 I g. In addition, the adsorptive potential distribution has been broadened. If we consider the central peak in these distributions, between say 50 to 60 K, to represent the graphite basal plane, we see that the fraction of basal plane has been reduced in the ground material, which indeed gives superior anode performance.
159
7.4 Results for "Nonporous" Carbons
KS75
(a)
Area =6.28 m2/g
20
(b)
40 60 80 Adsorptive potential (K)
100
120
KS75KM
6........----------------------. Area = 25.76 m2/g 5
O-+----~
o
20
40 60 80 Adsorptive potential (K)
100
120
Figure ].6 (a) The adsorptive potential distribution of a natural, low surface area graphite. (b) The same material after a fine-grinding procedure, showing a slightly broadened distribution and increased surface area.
7.4.3 Carbon Blacks An example of a much more heterogeneous surface is shown in Fig. 7.7. The adsorbent in this case is a carbon black designated C4, used by ASTM committee D24 as a reference reinforcing black. Again we see that the MNLDFT models provide an excellent fit to the adsorption data. The total surface area by the present method is 138.69 m 2 /g; the BET method gives 129.63 m 2 /g. The weighted mean value ofXsf/k is 53.61 K. The central mode of the distribution is seen to be somewhat lower than that for the graphites.
Chapter 7 The Surface Heterogeneity of Carbon and Its Assessment
160
(b) 12 - r - - - - - - - - - - - - - - - - ,
(a) 50 • -
Experimental data Fitted MNLDFT models
-
~
10
5
m
8
(ij (ij
'E Q) E Q)
~
6
4
2
0-+---r------,r-----r---r----,--~__1
1e-7 1e-6 1e-5 1e-4 1e-3 1e-2 1e-1
Relative pressure
1e+0
o -+-----r--
o
20
40
60
80
100
120
Adsorptive potential (K)
Figure 7-7 (a) A comparison of experimental data for nitrogen adsorbed at 77 K on ASTM carbon black C4 (points) with the fit given by the modified nonlocal density functional theory (MNLDFT) models (line). (b) The adsorptive potential distribution for the carbon black.
7.5
ACTIVATED CARBONS
The carbons discussed up to this point display a range of adsorptive potentials created chiefly by their surface roughness and chemistry; thus, their isotherms can be quite accurately modeled by a system of free, homogeneous surfaces of varying adsorptive potential. Their heterogeneity is then described by the area distribution of those potentials. Activated carbons cannot be energetically characterized by this method. These materials have a much more complex structure, providing many possible sources of energetic heterogeneity. As in developing any characterization method, one wishes to define and use the simplest model that yields reasonable and useful results.
7-5-1 Assumed Structure Activated carbons are usually visualized as an assemblage of graphitic planes arranged in a near-parallel fashion, thus creating a microporous solid having approximately slit-like pores of molecular dimensions. The resulting overlapping wall potentials produce greatly enhanced adsorptive potentials, so one may argue that the energetic heterogeneity of the material is to a large extent controlled by the distribution of its pore widths rather than the detailed nature of the pore walls themselves. Several current characterization methods are based on this simple model. However, within the slit pore structure, the pore walls may be of different and varying thickness, from a single carbon layer to essentially graphitic (more than five layers) and can also be of varying lateral extent. The graphene planes within a wall unit may have some crystalline stacking order, i.e., as in
161
7.5 Activated Carbons
hexagonal (aba ... ) or rhombohedral (abca ... ) graphite or may be completely turbostratic, with no discernable relationship. In addition, a wall unit may carry certain functional groups, typically containing oxygen, nitrogen, or sulfur, on its surface or especially at its periphery. Locally, a number of wall units may be ordered in a parallel fashion, creating a domain having a slit pore structure. At longer range, the orientation of these locally ordered domains is probably uncorrelated, leading to the possibility of interdomain pore spaces of indeterminate geometry and with perhaps a larger average width than that of the slit pores within the more ordered domains. In addition, it is by no means clear that such a structure is totally rigid. That is, it is possible that dilation and/or contraction of domains may occur as a result of the pressure tensors within the pore system [21, 22]. While still greatly simplified, the above picture leads to several sources of differing adsorptive potential. In estimated order of importance, these are as follows: 1. 2. 3. 4. 5.
The The The The The
distribution of pore widths distribution of wall potentials distribution of wall unit and domain size or area form and distribution of interdomain porosity quantity of functional groups
At the time of this writing, commercially available software includes only the first of these.
7.5.2 Example Applications of the Simple Model The commonly used DFT-based methods for characterizing activated carbon assume that the pores are geometrically slits with smooth, unterminated graphitic walls of constant wall potential. The experimental data are then modeled as a system of homogeneous, confined slits of varying width. The energetic heterogeneity of the material is therefore completely expressed in terms of its distribution of pore widths. The integral equation of isothermal adsorption, Eqn (7.12), for the case of pore-size distribution can be written as the convolution
Q(P)
=
f dH q(p, H)j(H)
(7.13)
where Q(P) is the total quantity of adsorbate per gram of adsorbent at pressure p, q(p, H), the kernel function, describes the adsorption isotherm for an ideally homoporous material characterized by pore width H as the quantity of adsorbate per square meter of pore surface, and j(H) is the desired pore surface area distribution function with respect to H. The kernel function is calculated by DFT for a confined fluid [10, 14, 15] and Eqn (7.13) solved by the methods already discussed.
Chapter 7 The Surface Heterogeneity of Carbon and Its Assessment
162
Isotherms for argon at 87 K adsorbed on typical activated carbons are shown in Figs 7.8 and 7.9, along with the reconstructed isotherm resulting from the pore width distributions shown. While the fit to the data is satisfactory in both cases, inspection of the pore width distributions obtained for these and many other activated carbon samples reveals a disturbing similarity: they all show deep minima at regular multiples of the probe molecule diameter, particularly near 1 nm (3xXo) [21,23]. This can be traced to packing effects inherent in the kernel function models that seem to be missing in the real data. Figure 7.10 shows how the pore fluid density as calculated by DFT varies with pore width, with density maxima near the pore width distribution minima. (a)
~
ICJ)
400
.
350 -
(b) 0.06 Experimental data DFTfitted
300
~ Q)
E 0.04
"'E 250 ~
:::I
(5
U
Q)
e0
>
200
~
co
as 150
'E 0.02 Q) E
~
as :::I 0
0.03
0 C.
UJ
u
~
~ 0.05
'"E
100
~ 0
E
50
0.01 0.00
0 1e-7 1e-6 1e-5 1e-41e-3 1e-21e-11e+0 1e+1
10 Pore width (A)
Relative pressure
100
Figure 7.8
(a) A comparison of experimental data for argon adsorbed at 87 K on Carbosieve G activated carbon (points) with the fit given by the nonlocal density functional theory (NLDFT) models (line). (b) The pore width distribution for the carbon. (a)
700
f
600
'"E
500
l-
en
. -
(b) Experimental data NLDFT fitted isotherm
"0 Q)
"0
E
Q)
0.05
:::I
(5
400
>
0.04
Q)
0a. co
300
'E Q) E Q) U
200
(tj
::::J
0
E
E
(tj
~
Ci ;;-- 0.06 ~
~
.c 0en
0.07
100
~
0 1e-71e-61e-5 1e-4 1e-3 1e-2 1e-1 1e+0 1e+1
Relative pressure
0.03 0.02 0.01 0.00 10
100
Pore width (A)
Figure 7.9 (a) A comparison of experimental data for argon adsorbed at 87 K on activated carbon RH572 (points) with the fit given by the nonlocal density functional theory (NLDFT) models (line). (b) The pore width distribution for the carbon.
7.5 Activated Carbons
0.040
M E ()
~--------------------,
0.038
..........
-8
a «
\'" \'" '"
\ ~....
.
.. ' .;::
..... ·~h ..' h r .
~".
-9
....,6;
......&~
....
\ ~\"
0 (/)
1:'
\'"
~""
~;;
;~.-
-2, where < r > is the mean value of the radial coordinate [24].
193
9.2 Endohedral Adsorption 20 __---...---..,..--........--,,---....--.r-"":"'"---w
~
1
~,I
r:I)
."
I
Q:I
10
++~\ ! ++++
"
+
-10
----
+ ~++
I----~~-----L . -.....
.................
I
_-_ .. _-_ .. _..,I
I
I
-------- '--20 & . . . - - - - ' - -.......- -........- -.......- -.......0.4 0.2 0.0
.........
0.6
rlR
Figure 9.5 Potential energy (dashed curve) and simplified model potential (full curve) for a 3He atom in a tube of radius 0.5 nm. The energies of the ground state and lowest azimuthally and radially excited states are shown as horizontal lines along with the corresponding wave functions (squares, circles and crosses, respectively). (Adapted from Ref. [37].)
1.5
....----~---~---~-------,
---------;r---------20
0.5
-------;r--------------------10
0.0 1...-..----'--------'--------'--------' 0.0 1.0 2.0
T(K)
Figure 9.6 Heat capacity per atom, in units of Boltzmann's constant, as a function of T for a noninteracting gas of 4He atoms within a tube with R = 0.5 nm The low T limiting behavior is that of a lD gas, while the high T limit is that of a 2D gas. (Adapted from Re£ [24].)
One can study tubes of larger radius than those mentioned above. In such cases, adsorption occurs in a set of cylindrical shells. This kind of behavior has been explored as a general model of adsorption in porous media, using a wide variety of techniques [38-44]. For lack of space, we ignore such large-pore behavior in the remainder of this chapter.
194
9.2.2
Chapter 9 Adsorbed Gases in Bundles of Carbon Nanotubes: Theory and Simulation
Axial-Phase Transition
If the pore radius lies in an appropriate range, relative to (T, there can occur a so-called axial-phase transition. The term refers to capillary condensation in the case when a "shell" phase, bound to the wall, is gradually augmented by the appearance of a 1D "axial phase," localized near r = O. This transition is exemplified by the simulation results for H 2 in Fig. 9.7. One observes that at low chemical potential p., the film is localized at r = 0.39 nm, at close approach to the C cylindrical surface. As p., is increased above a threshold value, the axial phase appears rather abruptly. This behavior is demonstrated further in Fig. 9.8, which separates the total coverage into axial and shell contributions. At saturation, their ratio is about 6.9. Since this transition is essentially a configurational phenomenon, similar behavior is expected for the case of a classical gas. Figure 9.9 shows an adsorption isotherm in such a case, computed with a lattice gas model in which seven "shell" sites near the tube wall surround each axial site. The figure compares results from mean field (MF) theory and Monte Carlo calculations [31, 32]. The axial "transition" is spread over a range of p., in both calculations since the temperature is above that of the MFaxial-phase condensation. Although the MF results are qualitatively correct, they exhibit a spurious shell-filling transition at low T, seen in the figure as a coverage discontinuity near the reduced chemical potential p., * = 2.2. Similar spurious transitions arise in virtually all density functional studies of this problem, as these employ an MF approximation of the effective potential experienced by the particles; the effect of omitted fluctuations is particularly acute in 1D systems [42-44].
12
-
p,=-419 K p,=-354 K ---- p,=-322 K -_...............
:§' 10 'c
M
::J
.ci
~ ~
:.0 (lj
.0
e0.
~
"00
8
III
'~'I'
1"11\
.,1 \'_ 6
4
c
Q)
-.:~
~
i~t\\ ~"i '
0
'\,. \
2
0
)\
\\\
~" \ ~
c'\'o...._ 0
2
3
4
Radius (A)
Figure 9.7 Dependence on chemical potential JL of H 2 film density in a pore with R= O.7nm at 10K. (Adapted from Re£ [34].)
195
9.2 Endohedral Adsorption
0.04 0.3
0.03
~ denotes average over an ensemble of many realizations of the surface. The most general statistical information for a continuous stochastic process is given by its generating functional. If we assume that the AES is a Gaussian stochastic process, then its generating functional is given by [22, 23]:
F(a) == (exp {f d2 Ra(R) [E(R) - E]})
= exp [~f where H(R,
f d Rd R'a(R)H(R, R') a(R')] 2
2
(10.6)
R') = ([E(R) - E][E(R') - E]) is the covariance function and E =
(E(R»).
From the generating functional, the multivariate probability density distribution for the adsorptive energies at n points on the surface is obtained as:
where, by virtue of the condition (10.5), the covariance matrix (10.8)
10.3
2 17
Generalized Gaussian Model
is a function of the relative position vector between two points. Here n is the adsorptive energy dispersion and C the correlation function. If furthermore the surface is statistically isotropic, C is only a function of the distance r between two points. In this model the mean value of any macroscopic quantity of interest depending on the AES could then be evaluated by knowing E, 0, and C(r). The correlation function C(r) carries all the useful information about the energetic topography and should, in principle, be determined from the geometric and chemical structure of the adsorbent (even though the methodology to achieve this has not been developed so far). However, we could simplify the model even more by proposing for C(r) a simple Gaussian decay like:
(10.9)
where ro is the correlation length. This expression, which we do not intend to take as a realistic correlation function valid for any surface, simply stresses that the spatial correlation between adsorptive energies at points separated by a distance r < Yo is very high (close to 1) while for r > ro it is very low (close to 0). Thus the present model becomes very attractive in the sense that the energetic topography is characterized by a single parameter, the correlation length, and this opens the possibility for the determination of the three simple parameters of the model (E, fl, and ro) by, for example, fitting experimental adsorption isotherms. It is worthwhile to remark that the present model is a continuous one and not a lattice model of adsorption sites. This is an appealing feature, since, as we can see from Fig. 10.1, adsorption sites hardly form a regular lattice and furthermore many of them are so shallow that an adsorbed particle will most probably be quite mobile on appreciably large regions. To obtain a manageable equation for the adsorption isotherm in this model, without loosing the generality of a continuous model, we make use of a virial expansion for the two-dimensional spreading pressure 4J of the adsorbed phase [24]:
(10.10)
where p is the adsorbate surface density and Bn (1) is the nth two-dimensional virial coefficient. If the adsorbed phase is in equilibrium with an ideal gas phase whose density is Po and whose pressure is p, then making use of Gibbs equation Pod4J = pdp, the adsorption isotherm equation is given by:
p = K(1)pexp
[E
_n_
n~2 n-l
Bn (1) pn-l]
(10.11)
Chapter 10 Energetic Topography Effects
218
where K( 7) is an integration constant. By assuming that the potential energy of the system of adsorbed particles is the sum of the interparticle potential
Ugg
(IR
i -
R1 1), and the gas-solid potential [1]
(10.12) and that the stochastic process E(R) has the distribution given by Eqn. 10.7, the coefficients in Eqn. 10.11 are obtained as [12]: (10.13)
(10.14) where
S~,2 =112 = expl-Ugg (IR i - ~D IkB TJ -1 S~,2,3 = ftJtJ;3 S~,2,3,4
=ftJt~J;J;J;4 + 6ftJt~J;Jh4 + 3ftJ;.JhJt4 and so on.
It is clear that the calculation of gas-solid virial coefficients is very difficult, so that only the first few of them could be evaluated. This means that the model will be useful only at low values of the adsorbed phase density. But on the other hand, the most important effects of heterogeneity can be seen for the low-pressure part of the adsorption isotherm. In order to study how the first few virial coefficients depend on the energetic topography, we assume an interparticle interaction given by a LJ potential: (10.15) where (J is the particle diameter and kB Tgg is the depth of the potential. For the LJ potential, Eqn. 10.15, and introducing the notation E = -kB T a and n = kB T:, we obtain (see detailed calculations in Ref [25]): B2 (7)
= ~ + rg n { E
i [ -
(
~y exp ( -l (~r) ]-E
00
B3 (7)=- (2;)2
[10
00
tgt dt - 3 10
i [ -
(
~ r]}(10.16)
~~g2dt+310 ~lg~dt- 10 ~~dt] 00
00
(10.17)
10.3
21 9
Generalized Gaussian Model
where
gi(t) =
fa
b
drFi(r)Vo(tr)
(10.18)
r
F(r) = exp {4; [(~Y2 - (~r]} exp [ (~ e-H~)2] i
F2 (r)
= exp [ (~
'I' = 1T
fa
r
exp ( -liz
(~r)]
(10.19)
(10.20)
b
rFi (r)dr
(10.21)
and E i is the exponential integral function. The integrals involved in B 2 and B3 can be evaluated numerically. B4 could also be evaluated numerically within reasonably large computer time, but it would not be worth the much greater effort, because already at very low adsorbed phase density topography effects could be appreciated. Adimensional virial coefficients can be defined as B: = Bn/ (7TU 2 /2)n. The 2nd and 3rd coefficients are shown in Fig. 10.2 as a function of T/Tgg for Ts/Tgg = 2.0 (which represent a reasonably high heterogeneity with respect to interparticle interactions) and different values of roo As can be seen, the sensitivity of the virial coefficients with respect to the correlation length Yo is very high at low temperatures and is still appreciable even at a relatively
4
Ts/Tgg
=2.0
4
2
- 0 - '0=0 ____ '0 =0.20-
3
B;
'0=0-
0
B; -0-
----
-2
'0=
2
00
'0= 0 '0= 0.20-
---0- '0=0~
-4
'0=
00
0 0
(a)
2
3
4
T/Tgg
5
6
7
0
(b)
234
5
6
7
T/Tgg
Figure 10.2 Normalized gas-solid virial coefficients for a Lennard-Jones potential, as a function of the reduced temperature TITgg' for different values of the correlation length TO and for a given value of the standard deviation of the adsorptive potential kB T s ' Adapted from Re£ 25.
Chapter 10 Energetic Topography Effects
220
high temperature. As T:/Tgg decreases (figures not shown here), the effect of topography becomes weaker and practically disappears for T:/Tgg < 0.5. It is interesting to analyze the adsorption process to understand the peculiar behavior of B2 at low temperatures. For Yo = 0 (completely random topography) and Yo --+ 00 (macroscopic homogeneous patches) the relative positions of adsorbed particles are not dictated by the adsorption energy topography but rather by the interparticle potential, with prevalence of the attractive region, then making the integrand in Eqn. 10.14 preferentially positive, and therefore B2 --+ +00 as T --+ O. For 0 < Yo < 2(J, on the contrary, adsorbed particles are forced by the adsorptive energy topography to be close enough so that the repulsive part of the interparticle potential makes the prevailing contribution to the integrand in Eqn. 10.14, and B2 --+ -00 as T --+ O. As can be easily understood, the virial coefficients for Yo greater than a few particle diameters will behave approximately as for Yo = 00. Once the virial coefficients have been evaluated, the adsorption isotherm for low pressure is obtained through (10.22) (10.23) The constant K (T), known as the Henry's constant, representing the slope of the adsorption isotherm at a very low pressure, depends not only on the
TiTgg
0.3
--0-
-------D-
0.2
--atr-
= 2.0
'0 = 0 '0 =0.2a '0 = a '0 = 00
P
0.1
0.0
-+-~~~-r-r-"T""TTT~"'T""""""T""'T""TTT1rrr--,..-r-T"TTTI~""""""rTTTT-rr--T""""T""'1rTTTTrr-r-r-J"TTTT'Ii
1E-3
0.01
0.1
1
10
100
1000
10000
p/K(T)
Figure 10.3 Adsorption isotherms calculated from the GGM for Lennard-Jones interacting particles, for TITgg = 2.0 and different values of the correlation length. Adapted from Re£ 25.
10.4
Simulations on Ideal Heterogeneous Systems
221
mean adsorptive energy, E = -kB Ta, as classically believed [3], but also on the adsorptive energy dispersion 0 = kB Ts • Adsorption isotherms calculated from the above equations for Ts/Tgg = 2, T/ T gg = 2, and different values ofthe correlation length ro are shown in Fig. 10.3. The effect of the correlation length can be clearly appreciated as a considerable decrease in adsorption density as ro increases. Theoretical adsorption isotherms could be fitted to experimental ones obtaining the parameters K(T), ~, and ro, characterizing the AES for a given real gas-solid system. In what follows, however, we point to a quite stronger test of the GGM, namely, we produce artificial (computer-made) heterogeneous adsorbents with well-controlled energetic topography, determine the AED and the correlation function corresponding to the gas-solid system, then simulate the adsorption process in the continuum, and finally compare the observed behavior with the predictions (not data fitting) of the GGM.
10.4 SIMULATIONS ON IDEAL HETEROGENEOUS SYSTEMS A collection of solids is prepared as explained in Section 10.2, corresponding to different concentrations of impurity atoms, and their AES are generated. We can then study the statistical properties of these AES, like the AED and the spatial correlation function C(r). These statistical properties for a set of ideally prepared heterogeneous solids are shown in Figs. 10.4 and 10.5. As the concentration of impurity atoms increases, the mean value of the adsorption energy distribution (Fig. 10.4) shifts toward lower energy values (stronger adsorption) and its dispersion also increases. At the same time, the spatial correlation function (Fig. 10.5) presents an attenuated oscillatory behavior, with the decaying being slower for higher concentrations of impurity atoms. Once the ideal heterogeneous solids are prepared, the adsorption process is simulated through a continuum space Monte Carlo method in the grand canonical ensemble [26, 27]. The simulation method can briefly be outlined as follows. (a) A value of the pressure, p, and temperature, T, is fixed. (b) An arbitrary initial state with N adsorbed particles, S[y, is established (e.g., by adsorbing N particles at randomly chosen positions on the solid surface) and its energy is calculated as
(10.24)
222
Chapter 10 Energetic Topography Effects
-15.0
-12.5
-10.0
-7.5
-15.0
-12.5
-10.0
-7.5
U/Cgs
U/Cgs
(b)
(a)
Figure 10.4 Adsorptive energy distributions (AED) for ideal heterogeneous solids with different concentrations of impurity atoms. Adapted from Re£ 25.
-0--_____
'0=5.030 % '0 =3.5 70/0
--0--
0.8
'0=3.0 1 %
'0 =0.0
0%
0.4
0.0
-0.4
o
4
8
12
([A]
Figure 10.5 Comparison between "real" spatial correlation functions and those assumed by the GGM. Adapted from Re£ 25.
10.5
Comparison Test for the GGM
223
(c) One of the following three processes is randomly chosen with equal probabilities:
• Particle Displacement. A particle is chosen at random and a change in its position by a displacement vector Sis attempted. The modulus of the displacement vector is fixed but its direction is randomly chosen. The energy of the final state of the system (if the displacement were accepted), is calculated and the transition is accepted with probability
U(Sj),
W(S~ ---+ S;)
= min {1, exp [- (
U(Sj)k-TU(S["»)] }
(10.25)
B
• Particle Adsorption. A position on the surface is chosen at random and the adsorption of a new particle at that position is attempted. The transition is accepted with probability
m(S!'J~SN+l)=min{l I
'f
'
pA ex kB T( N + 1) P
[_(U(S[+l)-U(S["»)]} kB T
(10.26)
• Particle Desorption. An adsorbed particle is randomly chosen and its desorption is attempted. The transition is accepted with probability
W(S~ ---+ S;-l)
TN [(U(SN-l)-U(SN»)]} = min { 1, k~A exp 'f k T j
B
(10.27) where A is the area of the solid surface sample in the simulation. Step (c) is repeated until thermodynamical equilibrium is reached, and then further Monte Carlo steps (MCS) are executed to obtain the mean value of adsorbed particle density. By changing the value of p, the adsorption isotherm can be obtained.
10.5 COMPARISON TEST FOR THE
GGM
We now compare the predictions of the GGM with the behavior observed through simulations for the ideal heterogeneous systems As we can see from Fig. 10.4, the AED could be qualitatively described by a Gaussian distribution, as assumed by the GGM, whose dispersion increases as the concentration of impurity atoms increases. The case corresponding to 0 % concentration of impurity atoms is the less favorable, but it is also true that a
Chapter 10 Energetic Topography Effects
224
distortion of the AED in the high-energy region (weak adsorption energy) is not important for adsorption at low pressure, where the deeper adsorptive energy regions are preferentially occupied by adsorbed particles. It is to be expected that for more general heterogeneous solids, where heterogeneity could be due not only to impurity atoms but also to a number of defects, or even to the presence of amorphous structures, the AED would be even more similar to a Gaussian distribution. On the other hand, for the spatial correlation function, the Gaussian decay assumed by the GGM is also qualitatively acceptable, as can be seen from Fig. 10.5, where black symbols represent the Gaussian decay for different correlation lengths and the open symbols represent the spatial correlation function obtained from the AES for different concentrations of impurity atoms. In fact, even if the "real" correlation function presents the oscillatory structure induced by the periodic character of the solid lattice, these oscillations are not relevant to the adsorption of molecules, whose size is usually larger than the solid lattice spacing. What is important is the attenuation of the oscillations. Visual inspection of Fig. 10.1 suggests the importance of the size of the dark and bright regions, rather than the small grains within these regions. The important fact then is that the GGM provides a simple correlation function, which takes into account such a decay with only one parameter, the correlation length roo We now choose more or less appropriate (by visual comparison) AED and correlation length values for different samples of heterogeneous solids, and compare adsorption isotherms obtained by the GGM with simulated isotherms for those samples. This comparison is shown in Fig. 10.6, where black symbols represent simulated isotherms, whereas full curves represent GGM predictions. As
0.08
0.04
.: .'.•... 'a =2 '
. " ·.···'0=2.5 70/0
~~~~""""'L..--
1E-4
1E-3
1%
0.01
'a =3.5 30%
---+- 0.00
0.1
p[bar] Figure 10.6 Comparison between adsorption isotherms simulated on ideal heterogeneous solids (black symbols) and those predicted by the GGM (full lines), for three different samples. Adapted from Ref. 25.
10.6
Bivariate Model and Simulation Method
225
we have already mentioned, the comparison can only have significance at low pressure, given that we only use the virial expansion up to the 3rd coefficient. In this region, and considering that this is not the result of a parameter fitting procedure, we may say that the predictions of the model are satisfactory.
10.6 BIVARIATE MODEL AND SIMULATION METHOD
We now turn to a completely different kind of approach. We assume that the substrate is represented by a two-dimensional square lattice of M = L x L adsorption sites, with periodic boundary conditions. Each adsorption site can be either a "weak" site, with adsorptive energy E 1 , or a "strong" site, with adsorptive energy E 2 (E 1 < E 2 ). Weak and strong sites form patches of different geometry: (1) Square patches of size I (l = 1, 2, 3, ... ), which are spatially distributed either in a deterministic alternate way (chessboard topography), Fig. 10.7(a), or in a nonoverlapping random way (random topography), Fig. 10.7(b); (2) strips of transversal size I (l = 1, 2, 3, ... ), which are spatially distributed either in an
(a)
:::: ::::.... ........:::: .... ....:::: ............ ........:::: ........:::: .........:::: ... ....:::: :::: :::: :::: :::: .... ::::....::::....::::....:::: :::: :::: :::: . .:::: .... .... .. :::: .... :::: :::: ::::. . .... .. SSSS;;;;SSSS;;;;SSS
(c)
:::: ::::
mi .... .... ::::
.... :::: .... .... .... .... .... ....
:::: :::: ....
.... .... .... .... .... .... .... .... .... .... .... .... .... .... .... .... .... .... .... ....... ... ... ... ...... ... ... .... .... .... .... ..... ...
(b)
i;;;;iiis;;;;
.... .... .... .... .... .... .... ...... ... ... ..... ... ..... ... ...... ... ...... ... ... ...... ... ... ...... ... ... ...... ... ... ... .... .... . .... .... ... .... .... ... ....... ... .... .... ..... ... .... ... ..... ... .... ..... ... ...... ... ... ....... ... ... ...... ... ... .... .... ... ..... ...
(d)
.... .... ... ... .... .... .... .... .... .... . . ............ ............ ............ ............ ............ ............ ............ ............ :::::::: :::: :::::::: :::::::: ........ :::: ........ ........ ........ :::: ........ .... .... .... :::: . ........ . . ........ . . ........ . . ..... ... ............ .... :::::::::::: .... .... .... .... .... .... .... imim····imm~ .... .... . . .... .... . . .... . . ::::
.... .... .... .... .... .... .... .... .... .... .... .... :::: :::: :::: .... .... .... .... .... :::: .... .... .... .... :::: ....
............ :::::::::::: iiiiiiiiiiii :::::::::::: ............ ............ ............ ............ ............ :::::::::::: ............ ............ :::::::::::: ............ ............ ............ ............ ............ :::::::::::: ............ ::::::::::::
Figure 10.7 Schematic representation of heterogeneous bivariate surfaces with chessboard, (a), random square patches, (b), ordered strips, (c) and random strips, (d), topography. The patch size in this figure is 1== 4.
226
Chapter 10 Energetic Topography Effects
ordered alternate way, Fig. 10.7(c), or in a nonoverlapping random way (random topography), Fig. 10.7(d). In order to easily identify a given topography, we introduce the notation Ie for a chessboard topography of size 1 and, similarly, IR for random square patches, los for ordered strips, and IRS for random strips. Then, in Fig. 10.7(a)-(d), the topographies are 4e , 4R , 40s , and 4RS ' respectively. We also use the notation "bp" to refer to the extreme case ofbig patches topography (1-+ (0), i.e., a surface with one-half of weak sites and one-half of strong sites. The substrate is exposed to an ideal gas phase at temperature T and chemical potential J-L. Particles can be adsorbed on the substrate with the restriction of at most one adsorbed particle per site and we consider a nearest neighbor (NN) interaction energy w among them (we use the convention w > 0 for repulsive and w < 0 for attractive interactions). Then the adsorbed phase is characterized by the hamiltonian: H
= -M [ (8 1 01 + 82 ( 2 ) -
J-LO] + w
L ninj
(10.28)
(i,j)
where 0 = 01 + O2 is the total surface coverage (summing the coverages on weak and strong sites), ni is the site occupation number (=0 if empty or =1 if occupied) and the sum runs over all pairs of NN sites (iJ). Without any loss of generality, we can consider that all energies are measured in units of kB T, and that 8 1 = 0 and 8 2 = 8 1 + iiE, in such a way that the adsorptive energy is characterized by the single adimensional parameter iiE. The adsorption process is simulated through a Grand Canonical Ensemble Monte Carlo (GCEMC) method [26, 27]. For a given value of the temperature T and chemical potential j.L, an initial configuration with N = M /2 particles adsorbed at random positions is generated. Then an adsorption-desorption process is started, where a site is chosen at random and an attempt is made to change its occupancy state with probability given by the Metropolis rule, P = min {1, exp(-iiiHlkB T)}, where f1 iH = Hf - Hi is the difference between the hamiltonians of the final and initial states. A MCS is achieved when M sites have been tested to change its occupancy state. The approximation to thermodynamical equilibrium is monitored through the fluctuations in the number N of adsorbed particles; this is usually reached in 104 to 105 MCS. After that, mean values of thermodynamic quantities, like the surface coverage 0 and the internal energy U, are obtained by simple averages over m configurations. Then, o= < N > I M and U = < H > - J-L < N > where the brackets denote averages over statistically uncorrelated configurations. By changing the value of J-L, the adsorption isotherm at a given temperature can be obtained. Furthermore, from the simulation results, the differential heat of adsorption qd as a function of the coverage is calculated as qd = a < U> lao [28]. In our calculations we have used M ~ 104 and m = 105 . With this size of the lattice (L ~ 100, in such a way that it is a multiple of 1) we verified that finite size effects, which affect the isotherms in the case of repulsive interactions at much smaller sizes, are negligible.
10.7
Adsorption Results
227
10.7 ADSORPTION RESULTS We treat separately the cases of repulsive and attractive interactions. 10.7.1
Repulsive Interactions
Given that all energies are being measured in units of kB T, all results will be independent of the temperature and, furthermore, because the critical temperature for the appearance of a c(2 x 2) ordered phase in a zero-field Ising model is given by kB Te = 0.567w [29], there will be a critical NN interaction, we = 1.763668, above which the formation of the ordered phase is possible at (J = 0.5. In order to understand the basic phenomenology, we consider in the first place a chessboard topography with I = 4 (size of each homogeneous patch). Figure 10.8 shows the behavior of adsorption isotherms, (a), and qd((J), (b), for different square patches topographies for w = 4 and ~E = 24. It can be seen that all curves are contained between two limits: the one corresponding to Ie and the other corresponding to bp. For chessboard topographies, four different adsorption processes can be visualized, separated by shoulders in the adsorption isotherm and by steps in qd: (i) strong site patches are filled up first to (J = 0.25, where a c (2x2) structure is formed on them (in this region qd = 24); (ii) since 4w < ~E, the filling of strong site patches is completed up to (J = 0.5 (in this region qd decreases continuously from 24, zero-occupied NN, to 8, fouroccupied NN); processes (iii) and (iv), corresponding to the regions 0.5 < (J < 0.75 and 0.75 < () < 1, respectively, are equivalent to processes (i) and (ii) for
30
1.0 dE=24 0.8
20
w=4
0.6
---
10
()
--0- 3c
qd 0
----- 1R --.-2R
-10
---------- 3R ---._-- 4R
-----1 R f/" f;?:'___ 1c - - 2 R
0.4
. ~: .: ----I:s- 2c ---------- 3R - D - 3c ------- 4R
0.2
---.---- bp
--0- 4c -------- bp 0.0
-20 -30
-20
-10
0
10
20
0.0
0.2
0.4
0.6
0.8
1.0
()
{l
(a)
1c
- D - 2c
(b)
Figure 10.8 Adsorption isotherm, (a), and differential heat of adsorption, (b), for different topographies and repulsive interactions in regime I. Adapted from Re£ 30.
Chapter 10 Energetic Topography Effects
228
1.0
i1E= 12 -------- 1R
8
0.8
__ 2
w=4
R
------. 4 R
0.6
............. bp
o
()
0.4
-------- 1R
0.2
../(----- 1c - - 2 R --L::s- 2 -----. 4 R
_____ 1
-8
c --L::s- 2 c
c -D-4 ············· b p c
- D - 4c
-16
0.0 -20
-10
0
10
20
0.0
0.4
0.6
0.8
1.0
e
J-l (a)
0.2
(b)
Figure 10.9 Adsorption isotherm, (a), and differential heat of adsorption, (b), for different topographies and repulsive interactions in regime II. Adapted from Re£ 30.
weak site patches. Random topographies are seen to behave in a similar way with a particularly interesting feature: the behavior of a random topography of size 1 seems to approach that of a chessboard topography with an effective size Ieff > 1. As can be easily understood, as long as the condition w/ ~E :s 1/4 is satisfied, the adsorption process is similar to the one described above, i.e., strong site patches are filled first and weak site patches are filled after. We call this feature Regime 1. Figure 10.9 shows the behavior of adsorption isotherms, (a), and qd((}), (b), for different square patches topographies for w = 4 and ~E = 12. In this case, where w/ ~E 2: 1/3, the adsorption process follows a different regime, which we call Regime II: (i) the strong site patches are filled until the c(2 x 2) ordered phase is formed on them; (ii) the weak site patches are filled until the c(2 x 2) ordered phase is formed on them; (iii) the filling of the strong site patches is completed; (iv) the filling of the weak site patches is completed. It should be noticed that Regimes I and II are disconnected. In between, i.e. 1/4< w/ ~E < 1/3, the system behaves in a mixed transition regime changing continuously from one to another. Strip topography presents a similar behavior as square patches topography (not shown here), with the feature that ordered strips behave like chessboard square patches with a higher Ieff and random strips behave like random square patches also with a higher Ieff . A more detailed behavior of adsorption isotherms and differential heat of adsorption can be found in Refs [30, 31]. 10.7.2
Attractive Interactions
In the case of attractive interactions only Regime I is possible, i.e., for all values of ~E and w, strong patches fill first and weak patches fill last. Figures 10.10 and 10.11 show the typical behavior for square patches and
10.7
Adsorption Results
229
16
1.0
- e - bp
w=-1
0.8
- o - 4c
12
LiE=12
--0-- 2c
----*-
0.6 f)
qd 0.4
- - 4R - - - - 2R
0.2
- - - - .. 1R
1c
8
-e-bp - o - 4c
4
--0--2 c
w=-1 LiE=12
---*"-1 c
0.0
0 -18
-12
-6
0
0.0
0.2
0.4
J-l
0.6
0.8
1.0
f)
(a)
(b)
Figure 10.10 Adsorption isotherm, (a), and differential heat of adsorption, (b), for square patches topographies and attractive interactions. Adapted from Re£ 31.
1.0 16
w=-1 0.8
w=-1
12
LiE=12
LiE=12 0.6 f)
qd 0.4
-e-gp - 0 - 608
8
-0---2 08
-e-gp - 0 - 608
--*-1 08
4
-0---2 08
0.2
--*-1 08 0 0.0 -18
(a)
-12
-6
J-l
0
0.0
(b)
0.2
0.4
0.6
0.8
1.0
f)
Figure 10.11 Adsorption isotherm, (a), and differential heat of adsorption, (b), for strips topographies and attractive interactions. Adapted from Re£ 31.
for strips, respectively. In the last case only the ordered strips topography has been represented, as the density of curves is already high. The plateau in the isotherms and the corresponding abrupt drop in the differential heat of adsorption indicate that the strong patches are being filled before adsorption starts on the weak patches.
Chapter 10 Energetic Topography Effects
230
Again we observe that all curves vary between the bp topography and the Ie topography and that random topographies behave like the ordered ones with a larger effective size.
10.8 SCALING BEHAVIOR AND TEMPERATURE DEPENDENCE
The fact that both adsorption isotherm and heat of adsorption curves for different topographies, characterized by a length scale 1, vary between two extreme curves, suggests that we should search for some appropriate quantity to measure the deviation among these curves and study the behavior of such quantity as the length scale is varied. The quantity we found most suitable is the area between a given curve and a reference curve. For adsorption isotherms, this quantity, Xa' is defined as (10.29) where (JR(J-L) is the reference adsorption isotherm. A similar quantity, Xh' can be defined for adsorption heat curves. By taking as a reference curve the one corresponding to the bp topography, we obtain the plot of Xa as a function of 1 for different topographies corresponding to Regime I as shown in Fig. 10.12. Here we can see that Xa behaves as a power law in 1with an exponent a ~ - 2. Exactly the same behavior is also found for Xh' It is interesting to note that the
102 10°
101
L1E=12 w=-1
X 10-1 10-2
'0.
10°
X
..•...
·D.
.... ..•.. .... ....
10-1
D chessboard (5 = 1)
a
ordered strips (5=2)
. ~.
..•.
"0.
..•. ··A..
• random square patches (5= 2) .. random strips (5= 4)
(a)
....0.
Jeff
'0..
10-2
10-3
'0,
."
'0.
10-3
'0 .
··A.
".
..... D "'0.
..... D chessboard (5= 1)
a
10°
101
Jeff
'0.
. "'Q..
'0 ..
"Q..
.
'0 .
'".. '. 'a,. ..... li
ordered strips (5=2)
.
• random square patches (5=2) .. random strips (5=4)
(b)
Figure 10.12 Power law behavior of the quantity Xa showing the collapse of data for different topographies on a single curve when the effective length scale It1J is used: (a), repulsive interactions in Regime I; (b), attractive interactions.
10.8
Scaling Behavior and Temperature Dependence
231
exponent a is the same for repulsive (corresponding to Regime I) and attractive interactions and for all topographies, i.e., chessboard, random square patches, ordered strips, and random strips, as logarithmic plots are parallel. Straigthforward calculations demonstrate that the curves for X (either Xa or Xh) corresponding to the different topographies should collapse on the same curve as a function of an effective length 5cale (representing an effective patch size), leff' given by leff = 51, where 5 = 1 for chessboard topography, 5 = 2 for random square patches and for ordered strips, and 5 = 4 for random strips. The insets in Fig. 10.12 (a) and (b) show that this is indeed what happens. For repulsive interactions and for values of dE and w corresponding to Regime II, we find similar results, except that the exponent now has a different value, a ~ -3. Then, Xa behaves as a power law in the effective length scale, of the form In X = const + alnleff . This power law is valid over the whole range of energies, with different values of the exponent a. Figure 10.13 condenses the behavior of the scaling exponent for kB T = 1. We found that this behavior can be expressed as:
a a
= a l = -1.952±0.053;
for w/dE ~ 1/4
= a 2 + [12(1/3 - w/ dE)]I3(a l - ( 2 ); for 1/4 ~ w/ dE ~ 1/3 a = a 2 = -3.049±0.065; for w/dE ~ 1/3
(10.30)
with (3 = 0.42 ± 0.04 for repulsive interactions while a = a l = -1.9526 ± 0.053 for attractive interactions. As the temperature is changed, we find that the scaling exponent does not change for Regime I, whereas for Regime II its value tends toward that corresponding to Regime I as temperature increases [32]. This can be appreciated in Fig. 10.14(a), where hollow square symbols stand for a l and hollow circles -1.0
~-r--I----r----'--~-----"I---.:---r----rl--r---r-I-_
.-
Regime I --..'
:. -
Regime II
--..
-1.5 -
-
.. .
---ll------ft------~---- t, : : -2.5-2.0
a
• square patches, w> 0 -3.0 -
~ ~i~p~,e~:~hes, w < O ~----------Q-----------e------~
strips, W < 0
-3.5 -4.0
~
AE=4w
1
1
'\~
~/
AE=3w
-t---r--I----r---,I,...--+--,.I--I.....;....,----r'--r---r-,--f
0.1
0.2
0.3
0.4
0.5
Iwl/AE
Figure 10.1] Universal behavior of exponent a as a function of the adimensional variable willE for kB T= 1.
23 2
Chapter 10 Energetic Topography Effects
Regime I
Regime II
-2.0 ks T/!J.E= 1.20
-2
ks T/!J.E=0.33
-2.4
ks 77!J.E=0.16 ks T/!J.E=0.08
-3 -2.8
0.0
0.2
0.4
0.6
0.8
0.1
kBT/~E
(a)
0.2
0.3
0.4
0.5
w/~E
(b)
Figure 10.14 Dependence ofthe scaling exponent a on temperature. (a) Variation of a with k B T / LlE for Regime I (squares) and Regime II ( circles). Error bars represent Monte Carlo statistical errors. The solid line curve represents the fitting for Regime II given by Eqn. 10.31. (b) Overall behavior of a for different temperatures, represented through k B T / LlE. Curves for the intermediate regime are obtained by application of Eqn. 10.31 to Eqn. 10.30, while circles represent Monte Carlo results for wiLlE = 0.3. Adapted from Ref. 32.
symbols for a 2 • The full line represents a least squares fitting to the variation of a 2 given by
u 2 (kT/LlE) = -2-1.612exp(-S.2174kT/LlE)
(10.31)
If we assume this same variation for the values of the scaling exponent for the intermediate regime between Regimes I and II, we then obtain the general behavior represented in Fig. 10.14(b), where symbols in the intermediate regime correspond to Monte Carlo calculations. It is found that the scaling exponent a presents universality properties, in the sense that its behavior is identical for any value of LlE, for the different topographies considered, for different thermodynamical quantities (i.e., adsorption isotherm and differential heat of adsorption) and for different reference curves, even a theoretical one expressed, for example, through a mean field approximation for the bp topography like: (10.32) The corresponding reference curve for qd can be found by numerical differentiation through the general thermodynamical relation qd = (af.1/alnT)(} - kB T. This last universality property is extremely useful for practical applications, since if e 1 , e 2 , and w could be independently determined, as will be discussed
10.9
Conclusions
233
below, then the power law and the scaling exponent given by Eqns. 10.30 and 10.31 can be used to obtain Ieff from an experimental adsorption isotherm. These results suggest a method to solve the problem of the characterization of the energetic topography of heterogeneous substrates, which can be approximated by bivariate surfaces, through adsorption measurements of particles experimenting repulsive interactions. Adsorption measurements that are strictly necessary are the variation of the differential heat of adsorption as a function of coverage, qd (8), which can be obtained by using microcalorimetric techniques, and the adsorbate-adsorbate interaction energy, w, which can be obtained by low-energy electron diffraction (LEED) or scanning tunneling microscopy (STM) measurements at different temperatures to determine the critical temperature for the formation of the ordered c(2 x 2) structure. In the case of attractive interactions w can be estimated from adsorption measurements at very low pressures. With this, and since qd (0) = 8 2 and qd (1) = 8 1 + 4w, it is possible to determine 8 1 , 8 2 , and dE. Then, given the value of wi dE, the value of a can be obtained from Eqns 10.30 and 10.31. Finally, by choosing an appropriate theoretical approximation as a reference curve for qd (8), the value of Xh can be calculated allowing Ieff to be obtained from In X = const + aln leff' Note that the measurement of adsorption isotherms is not necessary for repulsive interactions, though it would be convenient to get an alternative value of Ieff to check the accuracy of the result.
10.9 CONCLUSIONS Several conclusions can be drawn from the present contribution. On the one hand, we have addressed the mobile adsorption (i.e., more suited to physical adsorption) of gases on heterogeneous surfaces at a low pressure. We have stressed the importance of the adsorptive energy topography, which can be taken into account by a theoretical model like the GGM, and we have extended such a model by calculating the 2nd and 3rd gas-solid virial coefficients for particles interacting through a LJ potential. The GGM turns out to be quite an attractive model due to its simplicity; in fact in this model the AES is statistically described by only three parameters: the mean value of the AED, kB T a , and its dispersion, kB I:, and the correlation length, roo The last parameter is the most relevant one describing the topography. The gas-solid virial coefficients were shown to depend strongly on the topography and, consequently, so does the adsorption isotherm at a low pressure. The only way to test the validity of such a model is to compare its predictions with the behavior of a system whose AED properties are well specified, and this is the case when adsorption is simulated on ideally constructed heterogeneous solids, as done here. The test turned out to be satisfactory for adsorption at a low pressure. From the above, we may say that the present form of the GGM can be used to fit experimental adsorption isotherms of physically adsorbed gases on heterogeneous solids at a low pressure,
Chapter 10 Energetic Topography Effects
234
obtaining in this way the parameters characterizing the heterogeneity. We may expect that the model would work better with substrates presenting a rough AES, because of either chemical impurities or roughness in the physical surface, such as in the case of activated carbons. Finally, since virial coefficients are found to be more sensitive to the correlation length at lower temperature, the appropriate adsorbates should be selected in such a way as to obtain experimental low-density adsorption isotherms at the lowest possible temperatures to ensure good sensitivity in the fitting parameters. On the other hand, we have studied by Monte Carlo simulations the adsorption of particles, interacting through a NN interaction w, on heterogeneous bivariate surfaces characterized by different energetic topographies. The heterogeneity is determined by two parameters: the difference of adsorptive energy between strong and weak sites, fiE, and an effective correlation length, Ie£[, representing the length scale for homogeneous adsorptive patches. Unique scaling properties and power-law behavior have been established for relevant adsorption quantities, such as the adsorption isotherm and the differential heat of adsorption. Two distinct filling regimes, Regime I and Regime II, separated by an intermediate mixed regime, are clearly identified in the adsorption process. The scaling exponent a as a function of wi fiE is found to follow a universal behavior. Its value is constant with temperature for Regime I, whereas it increases with temperature for Regime II and the intermediate regime toward the value corresponding to Regime I. This temperature dependence is given as an empirical equation obtained by Monte Carlo data fitting. These findings provide for the first time a method to characterize the energetic topography (i.e., obtain the parameters from experimental measurements) of a class of heterogeneous surfaces that can be approximately represented as bivariate surfaces.
ACKNOWLEDGMENTS
We gratefully acknowledge financial support from CONICET of Argentina and CONACYT of Mexico, which made possible the development of the present research.
REFERENCES
1. Steele, W.A. (1974). The Interaction of Gases with Solid Suifaces. Pergamon. 2. Jaroniec, M. and Madey, R. (1988). Physical Adsorption on Heterogeneous Suifaces. Elsevier.
References
235
3. Rudzinski, W. and Everett, D.H. (1992). Adsorption of Gases on Heterogeneous Surfaces. Academic Press. 4. Rudzinski, W., Steele, W.A., and Zgrablich, G. (1997). (eds). Equilibria and Dynamics of Gas Adsorption on Heterogeneous Solid Surfaces. Elsevier. 5. Jaroniec, M. and Brauer, P. (1985). Recent progress in determination of energetic heterogeneity of solids from adsorption data. Surf. Sci. Rep., 6, 65. 6. Riccardo, J.L., Chade, M.A., Pereyra, V.D., and Zgrablich, G. (1992). Adsorption and surface diffusion on generalized heterogeneous surfaces. Langmuir, 8, 1518. 7. Zgrablich, G., Mayagoitia, V., Rojas, F., et al. (1996). Molecular processes on heterogeneous solid surfaces. Langmuir, 12, 129. 8. Zgrablich, G., Zuppa, C., Ciacera, M., et al. (1996). The effect of energetic topography on the structure of the adsorbate. Surf. Sci., 356, 257. 9. Bulnes, F., Nieto, F., Pereyra, V., et al. (1999). Energetic topography effects on surface diffusion. Langmuir, 15, 5990. 10. Bulnes, F., Pereyra, V., Riccardo, J.L., and Zgrablich, G. (1999). Effects of the heterogeneous energetic topography on the collective motion of adsorbed particles. ]. Chem. Phys., 111, 1. 11. Gargiulo, V., Sales, J.L., Ciacera, M., and Zgrablich, G. (2002). Characterization of energetic topography of heterogeneous surfaces through the analysis of thermal desorption spectra. Surf. Sci., 501, 282. 12. Ripa, P. and Zgrablich, G. (1975). Effect of the potential correlation function on the physical adsorption on heterogeneous substrates.]. Phys. Chem., 79, 2118. 13. Steele, W.A. (1999). The supersite approach to adsorption on heterogeneous surfaces. Langmuir, 15, 6083. 14. Yang, M.X., Gracias, D.H., Jacobs, P.W., and Somorjai, G. (1998). Lithographic fabrication of model systems in heterogeneous catalysis and surface science studies. Langmuir, 14, 1458. 15. Lopinski, G.P., Wayner, D.D.M., and Wolkow, R.A. (2000). Self-directed growth of molecular nanostructures on silicon. Nature, 406, 48. 16. Fishlock, T.W., Pethica, J.B., and Eydell, R.G. (2000). Observation of a nanoscale chessboard superstructure in the Br-Cu (100) adsorbate system. Surf. Sci., 445, L47. 17. Nitta, T., Kuro-oka, M., and Katayama, T. (1984). An adsorption isotherm of multi-site occupance model for heterogeneous surface.]. Chem. Eng.]pn., 17,45. 18. Balazs, A.C., Gempe, M.C., and Zhou, Z. (1991). Polymer adsorption on chemically heterogeneous substrates. Macromolecules, 24, 4918. 19. Patrykiejew, A. (1993). Monte Carlo study ofadsorption on heterogeneous surfaces: finite size and boundary effects in localized monolayers. Langmuir, 9, 2562. 20. Nitta, T., Kiriyama, H., Shigeta, T. (1997). Monte Carlo simulation study for adsorption of dimers on random heterogeneous surfaces. Langmuir, 13, 903. 21. Nieto, F. and Uebing, C. (1998). Diffusion of adsorbates on random alloy surfaces. Eur. Phys. ]., Bl, 523. 22. Feller, W. (1971). An Introduction to Probability Theory and Its Applications, Vol. II, 2nd edn. Wiley. 23. Gardiner, C.W. (1985). Handbook of Stochastic Methods, 2nd edn. Springer. 24. Hill, T.L. (1956). Statistical Mechanics. McGraw-Hill. 25. Nazzarro, M. and Zgrablich, G. (2003). Energetic topography effects on mobile adsorption on heterogeneous surfaces at low coverage. Langmuir, 19, 6737. 26. Binder, K. (1986). Monte Carlo Methods in Statistical Physics. Springer-Verlag.
Chapter 10 Energetic Topography Effects
27. Nicholson, D. and Parsonage, N.G. (1982). Computer Simulation and the Statistical Mechanics of Adsorption. Academic Press. 28. Bakaev, V. and Steele, W.A. (1992). Grand canonical ensemble computer simulation of adsorption of argon on a heterogeneous surface. Langmuir, 8, 148. 29. Yeomans, J.M. (1992). Statistical Mechanics of Phase Transitions. Clarendon Press. 30. Bulnes, F., Ramirez-Pastor, A.J., and Zgrablich, G. (2001). Scaling behavior in adsorption on bivariate surfaces and the determination of energetic topography. J. Chern. Phys., 115, 1513. 31. Bulnes, F., Ramirez-Pastor, A.J., and Zgrablich, G. (2002). Scaling laws in adsorption on bivariate surfaces. Phys. Rev. E, 65, 31603. 32. Roma, F., Bulnes, F., Ramirez-Pastor, A.J., and Zgrablich, G. (2003). Temperature dependence of scaling laws in adsorption on bivariate surfaces. J. Phys. Chern., 5,3694.
POROUS TEXTURE CHARACTERIZATION FROM GAS-SOLID ADSORPTION Duong D. Do, Eugene A. Ustinov, and Ha D. Do School
of Engineering,
University of Queensland, St Lucia, Qld, Australia
Contents 11.1 11.2 11.3 11.4 11.5 11.6 11.7
Introduction Potential Models Classical Methods for Pore Characterization Density Functional Theory Monte Carlo Simulations Additional Features Conclusions Acknowledgment References
239 24 0 246 253 257 262 263 264 264
11.1 INTRODUCTION
Characterization of porous activated carbon and its derivatives has been a subject of great interest for many decades. Various tools for equilibria characterization are available in the literature, and they can be broadly classified into two groups: One is based on classical approaches while the other has firm foundation on molecular interaction calculations. Scientists constantly develop new tools or refine existing methods to better characterize porous carbons as the structure has significant effects on equili~ria as well as kinetics. Adsorption by Carbons ISBN: 978-0-08-044464-2
© 2008 Elsevier Ltd. All rights reserved.
239
Chapter
11.1.1
11
Porous Texture Characterization from Gas-Solid Adsorption
Carbon Structure
Carbon-based materials usually have a bimodal pore size distribution with one dominant peak being less than approximately 2 nm and the other major peak usually greater than 50 nm. The classification of pore size established by IUPAC [1] reflects neatly the range of pore sizes exhibited by carbon-based materials, which is micropores (pore size less than 2 nm), macropores (greater than 50 nm), and mesopores (falling between micropores and macropores). Micropores in activated carbon are dominantly used for storage of adsorbed molecules. The potentials exerted by the confinement of small pores are so great such that molecules inside those pores are not free from the attractive forces exerted by both walls of the pore [2]. Micropores are usually modeled as slit pores although this is a gross idealization of real pores, which are known to be finite, contain functional groups, defects, and do not have perfectly flat graphite surface [3]. Although there are attempts to relax some of the above restrictions, the ideal model of perfectly flat slit pore of infinite extent is still the most popular model used in almost all characterization methods. The complexity and the extreme computation time of more structured models are such that the ideal model of slit pore is still the obvious choice for pore characterization. Advanced carbon materials, such as carbon nanotubes and nanohorns, have pores of cylindrical shape and as such cylindrical pore is suitable for adsorption analysis for these types of materials. Since the discovery of carbon nanotube by Iijima [4], carbon nanotubes have been used by many as the candidate pores to study fundamentally the adsorption mechanism in cylindrical pores. In the past four decades, we have witnessed the significant development of various methods to describe microporous solids because of their important contribution to improving of adsorption capacity and separation. Various models of different complexity have been developed [5]. Some models have been simple with simple geometry, such as slit or cylinder, while some are more structured such as the disk model of Segarra and Glandt [6]. Recently, there has been great interest in using the reverse Monte Carlo (Me) simulation to reconstruct the carbon structure, which produces the desired properties, such as the surface area and pore volume [7, 8]. Much effort has been spent on studies of characterization of porous media [9-15]. In this chapter we will briefly review the classical approaches that still bear some impact on pore characterization, and concentrate on the advanced tools of density functional theory (DFT) and MC, which currently have wide applications in many systems.
11.2 POTENTIAL MODELS
The success of various models rests on the correct choice of the pairwise potential energy equation. In this section we will address the potential equations commonly employed for adsorbates used in pore characterization.
11.2
Potential Models
11.2.1
Fluid-Fluid Potential Models
There are many potential models that have been proposed in the literature. Among the popular ones that are currently enjoying widespread applications are the Lennard-Jones (LJ 12-6) equation and the Buckingham Exp-6 equation. The parameters of these equations are usually obtained by matching the theory (i.e., DFT) or simulation results (e.g., MC simulations) against various experimental properties, e.g., second virial coefficient, viscosity, vapor pressure, saturated liquid density, or surface tension, at the temperature at which the adsorption is carried out. Depending on the complex structure of the adsorbate molecule, simple atoms or spherical molecules can be assumed to behave as one-center interaction particle, i.e., they contain only one interaction site that involves in the interaction with the other atoms or molecules. Some adsorbates such as nitrogen and carbon dioxide contain more than one interaction site on each molecule. 11.2.1.1
Single interaction site particle
When a particle contains only one interaction site, the interaction between it and another is calculated with an equation that relates the interaction potential energy and the distance between two particles. One such equation is the LJ 12-6 equation, which contains two parameters, the collision diameter (J and the well depth of the potential B: (11.1 ) The molecular parameters of common adsorbates used in the pore structure characterization are listed in Table 11.1. It should be noted that these values are not unique as there are many other combinations of collision diameter and well depth of the interaction energy that have been determined in the literature [16]. Also noted in the table are the different sets of values that are used in DFT and in MC simulation. The difference is due to the mean field approximation assumed in the DFT analysis.
Table 11.1 Molecular parameters for simple molecules treated as 1C-LJ center
3.405 3.6154 3.685 3.81
119.8 101.5 164.41 148.1
3.305 3.5746 3.630 3.6177
118.05 93.746 163.1 146.91
Chapter
11
Porous Texture Characterization from Gas-Solid Adsorption
The LJ 12-6 potential equation is very popular, but that two-parameter equation is not flexible enough to handle many compounds adequately. This is resolved with equations involving more than two parameters, e.g., the Buckingham Exp-6 equation. This equation attracts interest from many workers [17-20] because it contains an additional parameter a that controls the steepness of the repulsive part of the interaction potential energy profile. It has the following form, and parameters for some gases are listed in Table 11.2: Table 11.2
Ar
Nz CH 4
'P(r) = {
Parameters for the Buckingham Exp-6 equation
123.2 131.49 113.5 100 152.8 160.3
e 1-6ja
11.2.1.2
3.866 3.784 4.040 4.12 4.206 4.188
Hirschfelder et al. [16]
14 15 16.2 13.6 14 15
Hirschfelder et al. [16]
Jones and Gray [21] Hirschfelder et al. [16] Errington and Panagiotopoulos [20]
[(00
r)] -(r- )6}
{ -exp 6 a 1- a
rm
m
(11.2)
r
Multisite particles
In pore characterization of carbonaceous materials, nitrogen and carbon dioxide have been commonly used. Nitrogen is used because it is readily available, while carbon dioxide is used as a probing molecule for smaller pores because of its small linear dimension and it can be used at temperatures close to the ambient temperature. Because of their shape, we should consider each molecule as a particle composing of many interaction sites. Each site on one molecule will interact with all sites of another molecule. We write below the interaction energy between a site a on a molecule i with a site b on a molecule j with a LJ 12-6 equation. ~a.'b)
'PI,}
= 4e(a,b)
u(a,b)) _ _ 12 _ (u(a,b)) _ _ 6] (a,b) (a,b) [ ( r· . r· . I,}
(11.3)
I,}
The subscript is used for particle while the superscript is for site. Thus for a given intersite distance ri~;,b) to calculate the interaction energy 'P~~,b), we need to know the cross collision diameter a(a,b) and the cross well depth E(a,b). They can be determined by invoking the mixing rule due to Lorentz-Berthelot
11.2
Potential Models
243
J
(LB) , a(a,b) = [a(a,a) + a(b,b)] /2 and 8(a,b) = 8(a,a) 8(b,b). Knowing the site-site interaction, the interaction between two molecules is simply: M M 'Pi,j
= L...J L...J 'Pi,j(a,b) '"' '"'
(11.4)
a=lb=l
where M is the number of sites on each molecule. We have just addressed the interaction energy between two molecules where the interaction is due to a dispersive force. Although nitrogen and carbon dioxide have zero dipole moment, they both possess quadrupole, e.g., the quadrupole moments of nitrogen and carbon dioxide are -4.9 x 10- 40 and -14.9 x 10- 4°Cjm 2 , respectively. The effect of quadrupole can be accounted for in the intermolecular interaction by specifying the charges and their locations on each molecule. The interaction energy due to electrostatic force between a charge a on a molecule i and a charge b on a molecule j is determined via the Coulomb law of electrostatic interaction:
'Pq;i,j
ab
l
(a,b)
=
41T8
qi qj
o
.
(a,b) y..
(11.5)
l,}
where 8 0 is the permittivity of free space, yi:;,b) is the distance between two charges a and b on the molecules i and j, respectively, q~ is the value of the charge a on the molecule i and qJ is the value of the charge b on the molecule j. The electrostatic interaction between two molecules then takes the form with M q being the number of charges on the molecule: Mq Mq
'Pq; i,j
'"' '"'
= L...J L...J 'Pq;(a,b) i,j
(11.6)
a=lb=l
Nitrogen
Cracknell et al. [22] proposed a two LJ site and a four-charge model (M = 2 and M q = 4). The four charges lie on the molecular axis joining the centers of two nitrogen atoms and they are symmetrical with respect to the molecular~ center of mass. The distance between two positive charges of 0.373e is 1.694
A,
while that between two negative ~harges of -0.373e is 2.088 A.
The distance between the two LJ sites is 1.094 A, and the collision diameter and t?e well depth of the interaction energy for nitrogen atom are (J'(N,N) = 3.318 A and S(N,N) jk = 37.8K. Bottani and Bakaev [23] proposed a two LJ site and a three-charge model (M = 2 and M q = 3). One positive charge (0.910e) is at the center of the molecular axis joining the two centers of nitrogen atoms and the two syn:,metric negative charge (-0.405e) are on the same axis with a distance of 1.1 A from each other. The collision diameter and the well depth of the interaction energy
244
Chapter
11
Porous Texture Characterization from Gas-Solid Adsorption
~
of a nitrogen atom are a(N,N) = 3.32 A and e(N,N) /k = 36.4 K. The distance between the two LJ sites is the same as that between two negative charges (i.e., the charge is on the LJ site). This model is less computer-intensive than the Cracknell's model because of one less charge to compute the electrostatic interaction. Carbon dioxide The model proposed by Harris and Yung [24] for carbon dioxide is commonly used for pore characterization [25]. In this model, there are three LJ sites with charges centered on each site. The molecular parameters are given below:
= 2.757 A, B(C,C) /k = 28.129 K a(O,O) = 3.033 A, B(O,O) / k = 80.507 K f, = 1.149A; qC = 0.6512e; qO = -0.3256e a(C,C)
The parameter f, is the distance between the oxygen LJ site and the carbon LJ site. 11.2.2 Solid-Fluid Potential Energy
The solid-fluid potential energy can be calculated by performing a summation of pairwise interaction between all the sites on an adsorbate molecule with all the atoms on the surface. This corrugated surface is important if the collision diame;er of the adsorbate molecule is comparable to the carbon-carbon distance (1.21 A) on the graphite surface or if the temperature is very low, when the structural behavior of the contact layer is very sensitive to this effect. However, for adsorbates having large collision diameter and high temperature, the assumption of structureless surface is reasonable and the surface can be assumed to be a continuum and the solid-fluid potential energy can be obtained by simple integration. 11.2.2.1 Slit shape pore
In the case of a single atom, its interaction with a structureless homogeneous surface made-up by a number of graphite layers, can be calculated from the Steele 10-4-3 equation [26, 27]: u
z - 41TB
sEC ) -
2
sfPs
1 a sf 10 1 a sf 4 a sf 4] [ 5 ( - z ) - -2 ( - z ) - 6.i(z + O.61.i)3
a Ll sf
(11.7)
where Ps is the density of the carbon center (114 x 1027 / m 3 ) , Ll is the interlayer graphite spacing (3.35 x 10- 10 m), and a sf and B sf are fluid-solid molecular parameters. The variable z is the distance between the atom and the plane
11.2
245
Potential Models
passing through the centers of all atoms of the outermost layer of the pore wall. The solid-fluid molecular parameters are usually obtained by matching the following theoretical Henry constant against the adsorption data on nonporous graphitized thermal carbon black: K
=
1 kT
/00 {exp [U Z) ] - 1} dz ----;;y Sf
(
(11.8)
o
where K = f/P. Here f is the surface excess. If the Henry constant is not available experimentally, the fluid-solid molecular parameters can be estimated from the usual LB rule. For carbon, the following parameters are commonly used u ss = 0.34nm and sssik = 28K. Equation (11.7) is the fluid-solid interaction energy for either atoms such as noble gases or lC-LJ molecules. For a polyatomic molecule with M centers of LJ type, the solid-fluid interaction energy can be determined the same way as we have presented earlier for fluid-fluid interaction. The interaction potential energy between a site a of the molecule i and the homogeneous flat solid substrate is calculated by the same 10-4-3 Steele potential [26, 27]: qJ(a)
= 41TP
I,S
e(a,s) [ a(a,s)Y C
Ll
1 (u(a,s)) 10 1 (u(a,s)) 4 [U(a,s)]4 } _ _ _ _ _ _ ----5 z 2 Z 6~(0.61~+z)3
1
(11.9) Knowing the interaction potential energy of the site a of the molecule i with the surface as given above, the solid-fluid interaction energy of the molecule i is 'Pi,s = L~l 'P~~. Once the solid-fluid potential energy for one wall is obtained, the potential energy between one molecule with a pore of slit shape and a width H is obtained from 'Pi, s(Z) + 'Pi , (H - z) 5
11.2.2.2
Cylindrical pores
The solid-fluid potential dealt with in the last section is for slit pores, and therefore it is applicable for solids such as activated carbon and activated carbon fibers. In the case of cylinder such as carbon nanotube, the interaction energy between a site a and the solid composing of Z concentric tubes is calculated from [28]:
'P~a) I,S
z
= 41TP C s(a,s) "L
{[u(a,s)]12 I n,6 _ [u(a,s)]6 I n,3 }
(11.10)
n=l
where I n,3 and I n,6 are calculated from the following integrals: (11.11)
Chapter 11 Porous Texture Characterization from Gas-Solid Adsorption
for m = 3 and 6. Here r is the distance of the interaction site a from the center of the tube. The variable an is the radius of the nth concentric shell, i.e., an = R + nli, where Ii is the spacing between two concentric shells, and R is the radius of the innermost carbon shell. This pore radius is defined as the distance from the pore center to the circular ring passing through all the carbon centers of the inner most shell. The integral of Eqn (11.11) can be expressed in terms of the hypergeometric function [29]. Knowing the interaction potential energy of the site a of the molecule i with the cylindrical pore as given in Eqn (11.10), the solid-fluid interaction energy of the molecule i and the pore is then calculated from 'Pi,s = L~l 'P~~. This potential energy equation has been used by a number of authors [30-34] in their analysis of solids having cylindrical pores, such as carbon nanotube and MCM-41.
11.3 CLASSICAL METHODS FOR PORE CHARACTERIZATION
Before discussing the two advanced methods for pore characterization, we would like to note that classical methods presented in the literature are applicable to mesoporous solids [35-38]. Among the early methods for characterizing microporous solids is the Hovarth-Kawazoe method [39] and it was later modified by a number of authors [40-44]. 11.3.1
Barrett, Joyner, and Halenda Method
The method devised by Barrett, Joyner, and Halenda (BJH) [35] is one of the earliest methods developed to address the pore size distribution of mesoporous solids. This method assumes that adsorption in mesoporous solid (cylindrical pore is assumed) follows two sequential processes - building up ofadsorbed layer on the surface followed by a capillary condensation process. Karnaukhov and Kiselev [45] accounted for the curvature in the first process, but Bonnetain et al. [46] found that this improvement has little influence on the determination of pore size distribution. The second process is described by either the Cohan equation (for adsorption branch) or the Kelvin equation (for desorption branch). 11.3.2
Broekhoff-de Boer Method
Among many classical approaches available in the literature, a method developed by Broekhoff and de Boer (BdB) [47-53] for description of vapor adsorption and desorption in cylindrical pores and slit pores is the most thermodynamically rigorous and elegant for more than 35 years. This method relies on a reference system, which is a flat surface having the same structure and surface chemistry as that of the adsorbent. The pores of the adsorbent can have either
11.3
Classical Methods for Pore Characterization
247
a cylindrical shape or slit shape. Their theoretical analysis rests on the following assumptions: • The adsorbed phase has the form of a liquid film whose density is equal to that of saturated bulk liquid. The liquid film-vapor interface is of zero extent. • The contribution of gas-like phase to the amount adsorbed is neglected. • The surface tension of the liquid film is the same as that for the macroscopic liquid and does not depend on the film thickness and the interface curvature. • The solid-fluid potential varies with the distance from the flat surface the same way as from cylindrical surface regardless of the surface curvature and from the pore wall of slit pore. All these assumptions do not exactly agree with results obtained from molecular simulations. However, errors resulting from these assumptions in the case of cylindrical pore and in the case of the reference flat surface may partly compensate each other. The advantage of the BdB method is that in the framework of their model all thermodynamic derivations are strictly correct. Details of this method can be found in the excellent papers by Broekhoff and de Boer.
11.3.3 Dubinin Methods The Dubinin-Radushkevich (DR) equation was originally devised as an empirical expression of the Polanyi adsorption potential theory, and due to its simplicity it has been widely used to correlate adsorption data in many microporous solids despite its failure in giving the correct Henry constant at extremely low pressures. This equation is based on the premise that adsorption in micropores follows a mechanism of pore filling rather than the molecular layering and capillary condensation as proposed for mesoporous solids. It has the form:
() =
W Wo
= exp [_ (RTln po/ p )2] Eof3
(11.12)
where WI W O is the fraction of the micropore volume that is occupied by adsorbate molecules, f3 is called the similarity constant (benzene is chosen as a reference, i.e., f3 = 1) and Eo is the characteristic energy and is related to the mean micropore size. The DR equation describes reasonably well, adsorption data of many vapors in carbonaceous materials that have a wide pore size distribution (PSD). For fine microporous solids having narrow PSDs, the Dubinin and Astakhov (DA) equation was proposed by replacing the exponent 2 in Eqn (11.12) by n, where n is usually referred as the heterogeneity factor. This factor usually falls in the range of 1.5-3.0, and it can be as high as 5-6 for fine microporous solids such as zeolites. To describe solids having a distribution in terms of either energy or pore size, Stoeckli [54] and Huber et al. [55] proposed to use the DR equation as a local isotherm in an integral equation to correlate adsorption isotherm of
Chapter 11 Porous Texture Characterization from Gas-Solid Adsorption
heterogeneous solids. This distribution is in the form of the micropore volume with respect to the parameter B (B = 1/ Eo~). Stoeckli [56] described the distribution in terms of the pore size rather than energy. To achieve this, they related the characteristic energy Eo in terms of the micropore size as given below.
10.8
H=----
(11.13)
Eo -11.4
where H is in nm and Eo is in kJ/mol. Using the DA equation with n = 3 as the local isotherm, the overall isotherm equation can be written in terms of an integral involving the micropore size distribution in the integrand. Fitting this overall isotherm against experimental data, the parameters involved in the micropore size distribution can be optimally derived, from which the micropore size distribution can be deduced.
11.3.4 Horvath-Kawazoe Method and its Modifications The Horvath and Kawazoe (HK) method [39] was developed to determine the PSD of active carbons from nitrogen adsorption isotherm. All pores are assumed to have slit shape. This method rests on the assumption that the adsorption state of a pore is either empty or completely filled. The demarcation pressure between these two states is called the pore-filling pressure, and it is a function of pore width. The equilibrium of a pore exposed to a bulk phase of constant chemical potential is obtained from the minimization of the following grand thermodynamic potential:
n=
(11.14)
F-nJL
where F is the Helmholtz free energy, n is the number of molecules adsorbed in the pore, and JL is the chemical potential. If the pore is empty, the grand thermodynamic potential n is zero. When the pore is completely filled, the grand potential is a function of chemical potential. It is positive for low chemical potentials and become negative at higher chemical potentials. The chemical potential at which this grand potential changes is zero, is the pore-filling chemical potential. Thus the pore-filling chemical potential is simply equal to the molar Helmholtz free energy of the adsorbed phase, i.e., (11.15)
ILf = F
The molar Helmholtz free energy of the adsorbed phase is simply the sum of the intrinsic Helmholtz free energy and the solid-fluid potential averaged over the adsorbed phase. Assuming a liquid-like behavior of the adsorbed phase, this free energy is given by H-lTs[
-F= [G - -POv ] L M
1
+--H 2(Ts[
f
udz
(11.16)
11.3
249
Classical Methods for Pore Characterization
Here G L = J.L (1) + kB T In Po is the molecular Gibbs free energy of the bulk liquid at the saturation pressure Po; u is the solid-fluid potential at a distance z from the pore wall. The chemical potential J.Lf in the bulk phase at the filling pressure Pf is J.L (1) + kB TIn Pf· Then combining Eqns (11.15) and (11.16) yields the following basic equation relating the pore-filling pressure Pf vs pore width: 0
0
H-asf
kB TIn (Pf) Po
=
1
H -2usf
f
udz
(11.17)
asf
Note that the solid-fluid potential energy in the above integrand is a function of H (e.g., the 10-4-3 Steele potential). Let us illustrate the HK method in the case of nitrogen adsorption in carbon slit pores at 77.35 K. For this system, the potential well depth ssf / kB is 56 K and the solid-fluid collision diameter u sf is 0.3488 nm. The pore-filling pressure vs the pore width, obtained from Eqn (11.17), is shown in Fig. 11.1. The solid line is calculated by the nonlocal density functional theory (NLDFT), which will be described to some detail in Section 11.4. As seen in the figure, the curve obtained with the HK method (dashed line) correlates with NLDFT much better than that calculated with the Kelvin equation (dash-dotted line). Knowing the pore-filling pressure as a function of pore width, for a given bulk-phase pressure p, the width of the pore that is just instantly filled at this pressure is denoted as H f . All pores having widths smaller than H f will be filled while those having widths greater than H f are still empty. Thus the overall 100
.------------._-._-.---.---._-._-.---.---._-._-.-------.=-~.
10-1
----
/.".-.
t·
i
10-2 10-3
;
;
10-4
;
;
~ 10-5
;
0: 10~
~
l
. I
! I !I
10-7 10-8
I I
10-9
I I
I
10-10
10- 11
/
-+-----.,.----lI-~-____r__-_____r----r--___r_--_r__-___I
o
2
3
4
Pore width (nm)
Figure 11.1 Pore-filling pressure dependence on the pore width for nitrogen adsorption in carbon slit pore at 77.35 K. (Solid line) NLDFT. (Dashed line) Horvath-Kawazoe method. (Dash-dot line) Kelvin equation.
Chapter
11
Porous Texture Characterization from Gas-Solid Adsorption
amount adsorbed is simply the total volume of those pores having the width smaller than H f multiplied by the liquid density. Hf(P)
a=PL
f !(Ht)dHt
(11.18)
The comparative accuracy and simplicity were the reasons why the HK method has enjoyed its popularity. This method was further extended to cylindrical [57] and spherical pores [40]. However, the pore-filling pressure is insufficiently accurate to predict pore size distribution with the same accuracy as that obtained with rigorous molecular approaches like grand canonical Monte Carlo (GCMC) simulations and NLDFT. Various attempts to improve the HK method have been made in the literature. One of such attempts is the method developed by Dombrowski et al. [43], who proposed a "weighted" version of the HK approach. They were guided by results obtained by DFT, in which the density profile across the pore exhibits an oscillational behavior with a period roughly equal to one collision diameter. This makes their modified HK method DFT-dependent. Analogous attempt to improve the HK method was made by Rege and Yang [41]. They considered layering of the adsorbed molecules with the assumption that each layer only interacts with adjacent molecular layers. However, both the attempts for improvement of the HK model rest on the same assumption of step-like local isotherm. It is known that the pore-wall wetting precedes the capillary condensation, resulting in quite involved shape of local isotherms, which strongly depends on the pore size. This shortcoming of the HK model was recently overcome by more rigorous thermodynamic analysis of adsorption in slit carbonaceous pores accounting for the dependence of surface tension on the adsorbed film thickness [44].
11.3.5 Enhanced Potential Method of Do and Coworkers Adsorption in mesopores is traditionally characterized by a mechanism of two sequential processes (e.g., the BJH method). Many attempts have been made to extend the applicability of the classical approach to smaller pores [58, 59]. With this allowance the range of applicability of the Kelvin equation could be extended only moderately. In the attempt to deal with micropores or pores of all sizes using the semiclassical approach, Do [60] introduced a concept of enhanced layering and enhanced potential. In the method of Do and coworkers [60-70], the mechanism is proposed in that the adsorption occurs by two sequential processes: (i) molecular layering and (ii) pore filling. At first, this method sounds like the same method that has been used in the last 60 years for the description of adsorption in mesopores and macropores. So what are the differences here? The differences lie in the enhancement in the adsorption affinity (due to the overlapping of potential exerted by opposite surfaces) and in the enhanced pore pressure in the core (due to the long-range interaction ofthe solid-fluid potential). Details of this method can be found in Do and Do [65]. We will only brief it here.
11.3
Classical Methods for Pore Characterization
The pore pressure used in the calculation of the adsorbed film thickness is calculated from Pp = Pexp (
-aC{) )
(11.19)
kT p
where 'P p is the mean solid-fluid potential energy in the inner core region. The parameter a is introduced because of the approximate nature of that equation. The mean solid-fluid potential energy 'P p is obtained as an average of the solidfluid potential energy profile over the domain of the inner core, 0, that is not occupied by the adsorbed phase. H/2
- f 'Pp(z)dzj f 'Pp =
n
n
f
'Pp(z)dz
dz = ( zo+t ) H/2-t-z
(11.20)
0
The pore pressure is directly responsible for the molecular layering and the pore filling. Having described the pore pressure, we now address the molecular layering process. This process can be described by any appropriate equation. If there is no or weak fluid-fluid interaction, we can use the BET-type equation, while if the fluid-fluid interaction is strong we can use the modified Hill-de Boer equation as suggested by Do and Do [63] to calculate the adsorbed film thickness t. In these equations the affinity constant is a function of pore size and the pressure involved in those equations is the pore pressure. Now we turn to the pore-filling process. We argue that this process is governed by the following equation, which is similar in form to the modified Kelvin equation: RTln
'VV p ) = _ 2 I M ( Po (H/2 - t - zo)
(11.21)
The difference between the above equation and the modified Kelvin equation is the use of the pore pressure. Substituting the pore pressure of Eqn (11.19) into the above equation gives H/2
f 'Pp(z)dz - YVM R Tin (pP ) = _zo+_t _ (H/2 t zo) o a
(11.22)
For large pores (mesopores and macropores), the contribution of solid-fluid potential is negligible (the first term in the RHS) and the above equation is reduced to the modified Kelvin equation. On the other hand, for small pores of molecular dimension the overlapping of potentials exerted by the two opposite walls is such that the overlapped potential outweighs the surface tension effect
Chapter
11
Porous Texture Characterization from Gas-Solid Adsorption
1.2 1.0 0.8 d
0.6 0.4 0.2 0.0 0
4
2
Figure 11.2 The dependence of Q' on the reduced pore width.
(i.e., the second term on the RHS is negligible). We see that the pore-filling process in small pores is dictated by the enhanced potential. First, we apply this method to calculate the pore-filling pressure vs pore width, i.e., the pore at which the pore is filled with adsorbates. The necessary parameter in the estimation of the pore pressure is a. We obtain the dependence of this parameter on pore width by matching the pore-filling pressure obtained by our method with the results of DFT and GCMC, and Fig. 11.2 shows this dependence for nitrogen and argon. It is interesting to note that this dependence on pore width is independent of adsorbate. The dependence of the pore-filling pressure vs pore width is shown in Fig. 11.3, where we observe good agreement between this method and DFT and GCMC.
10-10
+---':'~--+-----+--~----i------~------l
o
10
20
30
Pore width (A)
Figure 11.3 Reduced pore-filling pressure vs pore width.
40
50
11.4
253
Density Functional Theory
This method has been tested against the GCMC simulation [65], and the derived PSDs and the fitting of adsorption isotherm agree well with those obtained with the GCMC. It has also been applied to various activated carbons and is tested against the DFT theory for activated carbon processing pores of different size [69]. It has been found that for standard activated carbon, this method agrees well with the DFT while in purely fine microporous activated carbon, the two methods show some deviations. It is worthwhile to mention here that the DFT also disagrees with the MC simulations in small pores containing less than two molecular layers [71].
11.4 DENSITY FUNCTIONAL THEORY 11.4.1
Introduction of DFT
Density functional theory is a powerful tool to study many phenomena in physical chemistry and chemical engineering. It was popularized in the early 1960s by a number of authors [72-74]. But it is not until the 1980s that this theory had found widespread applications in many interfacial problems. Capillary condensation in pore was systematically studied [75], and the first paper [76] applying this technique to the problem ofPSD determination of carbon particle appeared in 1989. This work used a local DFT, and it is now superseded by the NLDFT, which was developed by Tarazona and Evans [77-79]. This is the method that is now widely used in the characterization of pore size distribution. 11.4.1.1
The NLDFT method
Application of NLDFT to adsorption of fluids in porous media is usually carried out at constant temperature and pressure (constant chemical potential). The equilibrium state of the grand canonical ensemble corresponds to the minimum of the following thermodynamic grand potential:
n=
(11.23)
F- nJL
where n is the number of molecules in the pore and is obtained from the integration of the local density over the volume of the pore n = p(r)dr. Here p(r) is the local density expressed in molecules per unit volume. In a confined space of a pore the density and the thermodynamic functions such as the Helmholtz free energy are distributed over the pore space. Letf(r) be the molecular Helmholtz free energy. Then the total Helmholtz free energy of the fluid confined in the pore is
J
F
f
= p(r)f(r)dr
(11.24)
Chapter 11 Porous Texture Characterization from Gas-Solid Adsorption
254
The system is said to be at equilibrium when the grand thermodynamic potential is a minimum. To perform this minimization, we need to determine the molecular Helmholtz free energy, and this is the crucial part of the DFT method as we shall show below. The molecular Helmholtz free energy f(r) may be expressed as a sum of four contributions: • • • •
the the the the
ideal part J:d (r) = kB T [In (A 3p(r)) - 1] excess repulsive part !ex [p (r)] attractive part of fluid-fluid intermolecular interactions uint (r) external part of solid-fluid interactions uext (r)
Here p(r) is the smoothed density; and A is the thermal de Broglie wavelength. The repulsive part of the Helmholtz free energy is usually calculated by the Carnahan-Starling equation derived for the hard sphere fluid [80]:
4- 3-2 (1 _Yj)2 '
r (-)-k T 17- 17 Jex
YJ -
B
(11.25)
where dHS is the equivalent hard sphere diameter. As we have mentioned earlier, there are different recipes for calculating the smoothed density, but in the case of a single component system the most popular prescription is that proposed by Tarazona et al. [79]:
p(r)
= f p (r') w (Ir -
r'l ,
P(r'» dr'
(11.26)
It was assumed that the weighting function could be approximated by a power serIes W
(Ir - r'l , P(r')) =
W o (Ir
- r'l) + Wi (Ir - r'l) per) + W 2 (Ir - r'l) [p(r)]2 (11.27)
The attractive part of the Helmholtz free energy is calculated via mean field approximation:
. uillt(r) = "21
J4> (Ir - r'l)
p (r') dr'
(11.28)
where ¢(r) is the attractive potential of two molecules. The factor 1/2 is because each molecule accounts for one half of ¢(r). This potential is expressed by the Weeks-Chandler-Andersen rw'CA) scheme [81] -Sf['
4>(r)
=
4cff [
(~ff y2 - (~£f
rl
r < rm
rill < r < rc
0, Here sf[ is the potential well depth, and rm
= 2i / 6
(Tf[.
(11.29)
11.4
Density Functional Theory
255
Table 11.3 Molecular parameters for Ar and N2 determined from bulk properties and surface tension
Ar N2
4.2712 (4.2617) 4.3405 (4.3155)
0.3380 (0.3380) 0.3581 (0.3575)
0.3318 (0.3305) 0.3537 (0.3575)
116.93 (118.05) 98.09 (94.45)
0.0125
34944
0.00888
28693
The condition of minimum of the grand thermodynamic potential requires its functional derivative ao/ap to be equal to zero. In the case of the onedimensional task it yields:
JL = kB TIn (A 3 p(z)) +!ex [p(z)]
+
f
int
p (z')i:x [p (z')] 'P (z, z') dz' + 2u (z)
+ uext(z)
where f:x is the derivative of!ex with respect to the smoothed density ') (
'P z, z
=
W o (Iz
(11.30)
p, and
- z'l) + P(z') Wi (Iz - z'l) + [p (Z,)]2 W 2(Iz - z'l) 1_ PI (Z') - 2p (Z') PZ (z') (11.31)
The increase of the bulk pressure at a small increment after achievement of the equilibrium density distribution allows obtaining the adsorption branch of the isotherm. If the pore is wide enough, the capillary condensation will occur, with the pressure of the condensation being corresponded to the vapor-like spinodal point. Similarly, desorption branch of the isotherm will be obtained at the decrease of pressure. In this case, the capillary evaporation will occur at a liquid-like spinodal point. The equilibrium transition pressure is obtained by comparing the grand thermodynamic potentials corresponding to the adsorption and the desorption branches of the isotherm. It corresponds to the equality of these values of the grand thermodynamic potential. In Table 11.3 we present molecular parameters for argon and nitrogen, determined by the approach discussed in this section. In parentheses we present the values reported by N eimark et al. [31]. The surface tension and the liquid-phase density for Ar and N 2 at their boiling points, at which the molecular parameters were obtained, are also presented in this table. 11.4. 2
DFT Applications to Pores (Slit and Cylinder)
Application of the DFT theory to argon adsorption on graphitized carbon black at 87.29 K [82] is shown in Fig. 11.4, where the solid line is from the DFT theory.
Chapter
256
11
Porous Texture Characterization from Gas-Solid Adsorption
_100 C\I
N E 100 ::::::
(a)
E
..........
(5
E
-6 "'C Q)
-6 ~
60
..c
..c
"'C ct1
"'C ct1
C 20 :::J
'E
CJ)
:::J
o E
0
E 0 0.0
10
(;
0 CJ) 40
cd::
(b)
o E
80
cd:: 0.1 1'-.......-r-...........-rrrr--.....--.-.........-T"~---r--"r-T'"T'"'I"'T'TTT"""--.-""T"""'T""T"T"T'T'I'T'"
0.2
0.4
0.6
0.8
1.0
0.0001
0.001
p/Po
0.01
0.1
p/Po
Figure 11.4 Argon adsorption isotherm on graphitized carbon black at 87.29 K in linear scale (a) and logarithmic scale (b). (Solid lines) correlation by NLDFT. (Dashed lines) correlation with nonadditivity factor a of 0.0183. Specific surface area is taken to be 13.26m2 jg.
The common feature observed in both DFT and GCMC simulations is that these results overpredict the amount adsorbed in the reduced pressure region greater than about 0.2. This seems to indicate that the fluid-fluid interaction energy is overestimated in the presence of a solid surface, and therefore the usual assumption of pairwise additivity of fluid-fluid and solid-fluid potential energies is questionable. One way of resolving this issue is the application of the following quadratic equation for the potential of one molecule [83]: (11.32)
u=--------
kT
Here a is a positive parameter accounting for the multibody interaction. The dashed lines in the figure present the correlation by the NLDFT with the parameter a equal to 0.0183. This quite simple modification of NLDFT leads to excellent fitting of experimental data with parameters listed in Table 11.4. The parameters presented in this table may be used in modeling of adsorption in slit and cylindrical pores. For illustration, we show in Fig. 11.5 the local isotherms for nitrogen adsorption at 77.35 K in slit pores of various pore widths. As seen in this figure the shape of the local isotherm depends on the pore width. Having this information on local isotherms for a wide range of pore widths, they can be used to determine the pore size distribution.
Table 11.4 Molecular parameters for Ar and N2 adsorbed on graphitized carbon black [71]
Ar 87.3K N 2 77K
118.05 94.45
0.3305 0.3575
0.338 0.3575
58.01 56.10
0.3353 0.3488
0.0183 0.0242
11.5
257
Monte Carlo Simulations
40~---------------------------'
32 M
E
~ (5
E
24
S
~
.~
16
c
Q)
o
8
O~=-.....;!:~-=--..,..--=:::::::;:_:::;;;;;;;..p::::;;""iiiiiiijiillIIIIiiiiii~=;:'-_~
10-10
10-9
10-8
10-7
10-6
10-5
10-4
_ _~-=--:;::::::"":""---1
10-3
10-2
10-1
10°
p/Po Figure 11.5 Set of nitrogen adsorption isotherms in slit pores at 77.3 K. Dashed lines correspond to narrow pores having width H (from right to left), nm: 0.60, 0.62, 0.64, 0.66. In this range, the increase of the pore width shifts isotherms toward lower pressures. (Solid lines) the pore width (from left to right), nm: 0.68, 0.72, 0.76, 0.80, 0.84, 0.88, 0.94, 1.0, 1.1, 1.2, 1.4, 1.6,2,2.4,3, 4.
11.5 MONTE (ARLO SIMULATIONS 11.5.1
Ensembles Used in Simulations of Adsorption
Monte Carlo has been increasingly applied to solve many adsorption of interest. This is greatly due to the increasing speed of personal computer and the greater arsenal of MC simulation methods that have been developed in the past few decades. Among these methods, the GCMC and the Gibbs ensemble Monte Carlo (GEMC) are particularly useful for pore characterization. We will discuss briefly these methods. More detailed exposition can be found in many excellent books [84, 85]. 11.5.1.1
Grand canonical Monte Carlo
In the GCMC simulation [86, 87], we specify temperature, volume (pore volume), and the chemical potential in the simulation box. This ensemble is ideal to study adsorption where a solid adsorbent (or a single pore) is exposed to a bulk fluid of constant pressure or chemical potential. Like all MC simulation methods, a Markov chain of molecular configurations is produced. Any properties of interest can be derived by averaging over this Markov chain. In GCMC, there are three different moves used to generate the Markov chain which is then composed of a series of molecular configurations. They are (i) displacement, (ii) creation, and (iii) destruction. We briefly describe them. The first move is the displacement of particle. This can be done by choosing a particle in
Chapter
11
Porous Texture Characterization from Gas-Solid Adsorption
random. This particle is displaced to a new position, and the acceptance or , rejection of this move will follow the rule of acceptance commonly used in MC simulation: (11.33)
where d U is the difference between the configurational energy after the displacement and that before the displacement (dU = Unew - Uo1d ). For the calculation of interaction energy, the nearest periodic image convention [85] is used. The two remaining moves in the GCMC are the creation and destruction of a particle. They are selected with equal probability. In the creation move, a particle is created at a random position within the simulation box that already contains N particles. The newly inserted particle is denoted as the (N + 1)th particle. The insertion has the following probability of acceptance:
. {1 P = nun
V exp { [JL - U(N + 1) + U(N)]}} , A3(N + 1) kT
(11.34)
where V is the simulation box volume. In the destruction move a particle is selected in random and removed from the box. The selected particle is assigned as the Nth particle, with no loss of generality. The probability of such removal is
. {A
3
N exp {-[JL+U(N-1)-U(N)]}} P = ffiln 1, - V kT
(11.35)
The GCMC, in principle, is easy to apply and its extension to mixture is straightforward.
11.5.1.2
Gibbs ensemble Monte Carlo
Another method that is very useful in determining isotherm is the GEMC simulation [88]. This method was first developed for studying adsorption in cylindrical pores. It was later applied to study adsorption in slit pores [89]. In this method, the coexistence of two phases can be simulated without the need to establish the interface between them. As such the method can be used to study the vapor-liquid equilibria, and the low and high densities in a pore (phase transition in pores). Another advantage of the method is that, there is no need for the explicit determination of the free energy or chemical potential of the two phases. The MC steps are designed so that at equilibrium there will be equality between pressure, temperature, and chemical potential between the two phases. Because there is no need to consider the interface joining
11.5
Monte Carlo Simulations
259
the two phases, the two coexisting phases can be simulated in two separate simulation boxes. One phase is simulated in box I, and the other phase is in box II. The system of two boxes is considered such that the total volume, total number of particles, and temperature are remained constant during the course of simulation. In this GEMC, like the case of GCMC, there are three basic moves. The first move is the displacement of particle. This can be done by choosing a box in random (with equal probability between box I and box II) and then a particle in that box is chosen randomly. This particle is displaced to a new position, and the acceptance or rejection of this move will follow the usual rule of acceptance commonly used in MC simulation, as we have described before for GCMC. The second type of move is the interchange of particle between the two boxes. In this move, a box is selected in random, say box II, and then a particle is selected in random in this box and moved from this box to box I at a random position. This move has the probability of acceptance as given in Eqn (11.33) with d U being given by (11.36) where dUi (i = I, II) is the energy change that occurs in simulation box i, and N i and ~ are the number of particles in and the volume of the box i, respectively. The number of attempts to perform this move is such that the success in interchange is about 2%. The third move in the GEMC is the volume exchange. Let d V be the volume change. A box is chosen in random, say box I, and its volume is decreased by d V (i.e., the volume of box II is increased by d V to maintain constant total volume). The positions of all particles in those two boxes are scaled linearly according to the change in the linear dimensions of the two boxes. For example, if the box lengths of box I before and after the change in volume are ~old and ~new, respectively, then the x-positions of all particles after the volume change are (the same applies for the y- and z-positions) xjnew = xjo1d (z;ew / ~old). The probability of acceptance for this move is given as in Eqn (11.33) with dU being given by
We have described the three basic moves for the GEMC simulation. They are for spherical molecules. For complex molecules, we have an additional move to displacement, which is the orientation of the molecule. The GEMC just described can be used to study the phase equilibria, e.g., vapor-liquid equilibria or either pure fluids or mixtures. For phase equilibria in pore, this method has been applied to cylindrical pore [88]. This method basically involves the simulation of two simulation boxes. If the two boxes are volumes within the pore (called pore-pore GEMC), the method provides
Chapter 11 Porous Texture Characterization from Gas-Solid Adsorption
260
directly the densities of the two phases, if the phase transition exists. Unfortunately, this pore-pore GEMC does not provide the chemical potential at which the transition takes place. It can be found by either applying the Widom method or using the phase densities obtained from the pore-pore GEMC in the GCMC simulation curve. Instead of using the pore-pore GEMC, the pore-fluid GEMC can be carried out to determine the equilibrium between pore and bulk phase directly [90].
11.5.2 Monte Carlo Simulation for Slit Pores
The GCMC simulation can be readily performed for a set of pores of various widths of interest. The result will be a set of local isotherms. Of interest in pore characterization are the local isotherms for argon at 87.3 K and nitrogen at 77 K. The figures in Fig. 11.6 typically show the local "isotherms of argon at 87.3 K for slit pores having width 8, 10, 20, and 30 A. These isotherms are presented as pore density (kmol/m3 ) vs pressure (Pa). The pore density is defined as the number of moles divided by the available volume for adsorbate molecules.
1.0 M
M~ 0.8 \.. ,.:":,,,~ '.,'.' ': ,..
a 0.8
~
~c 0.6
~ 0.6 C
~
< .• ,....•
,.,+
,;, \
I'
; •... ,.,.., .,
, "."i
'·','·i.····,··.··,·.. ·.·.·,;,·· . ' .. ' , ; ,' , ~
'.,.•,.,
;
".
"""'."1
o
.'
0..
] 0.2 +.. ,.,.~,.".:+ ....
0.0 1~1~1~1~1~1~1~1~1~1~ Reduced pressure
v .'" , ....
't ...., .. ··'·'·:"1
."~"":';;'::'" ...,..;.,,'~+ ...,.. '.,';"'+ ..,. "''''''''l-'''' ";"", ,": ....,..,·'·'··"1
0.0 +--'-'--'+--~""'IF-'--'-'-+--'-~---'---'"-T--'--'-"""+--'-~~ 10-8 10-7 10-6 10-5 10-4 10-3 10-2 10-1 10° 101 Reduced pressure
.
1.0 J\.. t-'r
a
90°). When the contact angle is smaller than 90° it is very difficult to carry out a calorimetric experiments corresponding to spreading wetting because it is impossible to efficiently control any initial state where the powder would be unwetted, though in equilibrium with the saturating vapor:capillary condensation should occur between the particles.
12.2.3
Applications
Immersion energy is an integral quantity, which corresponds to the average interaction of the liquid with the entire solid. Each experiment only provides one figure, whereas, e.g., adsorption isotherms can discriminate between various kinds of successive interactions as the equilibrium pressure increases. Nevertheless, a careful analysis of the parameters contributing to the immersion energy allows to derive a most useful information about the solid surface. The immersion energy depends indeed on • The extent of the solid surface: for a given liquid-solid system, the immersion energy increases with the surface area (applications: measurement of the surface area either by comparison, using a reference material, or by applying a modified "absolute" Harkins and Jura method). • The chemical nature of the surface: for a given liquid, the immersion energy depends on the chemical nature of the surface: if the liquid is polar, the immersion energy increases with the polarity of surface chemical functions (applications: study of the influence of a heat treatment on the quality and amount of surface chemical functions, study of wettability). • The chemical nature of the immersion liquid: for a given surface, the immersion energy depends on the chemical nature of the liquid (applications: determination of the dipolar moment of surface sites by immersion in liquids of increasing polarities; analysis of the hydrophobic character).
12.2
Immersion Calorimetry of Carbons into Pure Liquids
• Porosity of the solid: if the solid is microporous, the molecules of the liquid may be too large to penetrate into all the pores (application: derivation of a micropore size distribution from the immersion energies in liquids of similar chemical nature but different molecular size). In the following paragraphs, we discuss the use of immersion calorimetry for the assessment of the surface chemistry, wettability, surface area and porosity of carbons. 12.2.3.1
Characterization of surface chemistry
The chemical nature of a solid determines its adsorptive and wetting properties. Now, the energy of immersion mainly depends on the surface chemistry but also, to some extent, on the nature of the bulk solid. For example, the interaction between water and silica has contributions from the bulk Si0 2 together with contributions from the silanol groups of the interface. Polar molecules are very sensitive to the local surface chemistry, whereas nonpolar molecules are more sensitive to the bulk composition. Interactions between a bulk liquid and a bulk solid through an interface are often described in terms of Hamaker constant [16]. Immersion calorimetry in apolar liquids was proposed to estimate the Hamaker constant [17]. The sensitivity of immersion calorimetry to the surface polarity has justified its use for characterising the surface sites. Dividing the energy of immersion into its various contributions leads to the following relationship: (12.11)
where E rep , stands for repulsive interactions at the interface, Ed is the contribution of dispersive forces (integrated over the entire volume) and Eo: is the energetic contribution of the polarity induced by the electric field at the interface. EJL is the contribution of the polar functional groups at the interface. It can be estimated from the average electric field at the interface F and the dipole moment JL of the liquid, as pointed out by Zettlemoyer et al. [18] or Morimoto and Suda [19]): (12.12)
where k is a constant which depends on the density of liquid molecules in the vicinity of the interface. This approach was validated by the nearly linear behavior observed when the immersion energy of a polar solid like titania is plotted as a function of polarity of the immersion liquid. It was also shown [18] that the immersion energy of a carbonaceous nonpolar surface is nearly independent of the immersion liquid. The slope allows one to calculate F, whereas the intercept at the origin provides the dispersive contribution. Numerous surface modifications were followed by immersion calorimetry. The energy of immersion and the kinetics of the process may help to distinguish between the removal ofphysisorbed water and the dehydroxylation as a function
284
Chapter 12 Porous Texture and Surface Characterization from Liquid-Solid Interactions
of outgassing temperature. Modifications making a carbon surface more or less hydrophobic were thus studied by immersion in water [20-22] but other polar molecules, like alcohols, were also useful to follow changes of polarity of carbon surfaces on oxidation [22-25]. Simple relationships were observed between the oxygen content, the acid-base properties and the immersion enthalpy of carbon surfaces in water [26-28]. The sensitivity to surface polarity also allowed to follow the regeneration of activated carbon surfaces [29]. In all above examples, the starting carbon was under vacuum. The result is averaged over all sites present on the surface. If information about the energetic distribution of surface sites is desired, it is necessary to carry out several immersion experiments after precovering the surface with the vapor of the immersion liquid up to various extents. This allowed Zettlemoyer et al. to plot "immersion isotherms", which are the fingerprint of the energetic distribution of the surface sites. Nevertheless, this is a time-consuming method which leads to the same information as that provided, in one experiment, by direct gas adsorption calorimetry, since the following equation holds [4]: (12.13) where d imm U( P) is the immersion energy after precoverage at the pressure P, d imm U the immersion energy under vacuum, uO" the molar surface excess energy and u1 the liquid molar energy. The difference uO" - u1 is assessed by gas adsorption calorimetry. In the case of porous solids, the main drawback of the immersion method is the filling of an unknown volume of pores during the precoverage step, unless the full adsorption isotherm of the vapor is previously known.
12.2.3.2
Characterization of wettability
Wettability is generally defined by the contact angle, which is the apparent result of the balance between interfacial free energies. Whereas it is relatively easy to "see it" and measure it on flat surfaces, its assessment on powders and porous solids is not straightforward. In the case of very hydrophobic porous solids (i.e., contact angles are >90°), it was shown by Gomez et al. [8] showed that both the pore size distribution and the contact angle can be assessed from a liquid intrusion experiment associated with calorimetry, like in the setup represented in Fig. 12.2. This approach is similar to mercury porosimetry, where the intrusion pressure and the intruded volume are continuously recorded, but, here, the extra measurement of the heat exchanged makes it possible, after appropriate correction for the compressibility of the liquid, to determine the variation of interfacial energy a,s the pore is progressively filled and therefore to evaluate the homogeneity of the surface. This interfacial energy only depends on the interfacial. tension and contact angle and it is involved in a process where
12.2
285
Immersion Calorimetry of Carbons into Pure Liquids
the solid-vapor interface is progressively replaced by a solid-liquid interface. For a reversible step, it can be shown [8] that
dU
=(T
dYlv
cos ( ) -
aT
d cos fJ + TYlv-- Ylv
aT
)
cos () dA
(12.14)
where () is the contact angle at equilibrium. Assuming that along the experimental intrusion path, be it reversible or not, the variation of interfacial energy is proportional to the wetted area, one can then plot the wetted area as a function of the pore size, which is a pore-size distribution curve. Its consistency was shown with the volume distribution obtained by applying the Washburn equation provided the (advancing) contact angle used for the calculation was constant. Also, the derived surface area compares reasonably well with the nitrogen-BET (BrunauerEmmett-Teller) surface area. This method therefore allows assessing, all at once, the pore size distribution, the contact angle and the homogeneity of the surface. If the advancing contact angle is lower than 90°, wetting is spontaneous inside the pores at a pressure equal or lower than the saturating pressure. Its measurement can be done by capillary rise. Nevertheless, this will only characterize the wettability of the external surface of the particles and not that of the internal surface of the pores. This is why, here again, calorimetric approaches were proposed to get an estimated value of the wettability in the case of powders. For example, Briant and Cuiec [30], showed that for a number of solid-liquid systems the following approximation holds: (12.15) This allows the ranking of a set of solids after their wettability by a given liquid, but only when the contact angle is zero. This approach was used to characterize the acid-base components of the surface tension [31, 32]. For values of the contact angle ranging from 0° to 90°, the method proposed by Spagnolo et al. [33] can be used. From the Young-Dupre equation and after integrating the Gibbs equation along the adsorption isotherm of the vapor, the following relations may indeed be derived:
-d imm U 1 -TIe - T [a (Ysl - Ys) / aT] cos () = - - - - - - - - - - - - - -
(12.16)
Ylv For solids of low energy, Spagnolo et al. [33] just keep the term which is evaluated to be 0.07 ± 0.02 from other sources [34]. Then cos () =
-~immu-0.07T
Ylv An equation of the form cos () =
-~immU
Ylv
+ TC
ays/aT, (12.17)
286
Chapter 12 Porous Texture and Surface Characterization from Liquid-Solid Interactions
where C is a constant, can be derived from Eqns (12.3) and (12.10) by assuming that the contact angle variation with temperature is independent of the nature of the system [10]. As stressed in the same paper, this equation is applicable only for systems with low value of immersion energy, which is precisely the case of carbons (and also, of course, of perfluorinated polymers). It should be pointed out that the methods used to evaluate wettability from calorimetric measurements have to be carefully used because they only lead to the energy part of the process and not to its free energy. Moreover, for contact angles above 90°, reliable experimental results, i.e., with full wetting, require the use of a high-pressure intrusion setup.
12.2.3.3 Characterization of surface area and porosity Theoretically, for a given chemical nature, the immersion energy of a nonporous solid should be proportional to the surface area and the corresponding coefficient should be available from a reference solid of known surface area. Nevertheless, the detailed surface composition and structure ofsolids with similar bulk composition and even crystallinity can be very different because of their chemical, mechanical or thermal history. Therefore, it would be unwise to use the relative measurement of surface area when both the surface and the immersion liquid are polar. Conversely, nonpolar liquids can be used for such a determination because the corresponding immersion energies are not sensitive to minute variations of the surface chemistry. This point will be addressed again later on in the case of microporous samples. Another way to derive a surface area is the Harkins and Jura "absolute method" [13], actually in its modified form by Partyka et al. [7]. The method is based on the "coating" of the solid particles by a water multilayer obtained at water saturating pressure and on the assumption that the liquid-vapor interface then surrounding each particle has the same area as the initially bare solid surface. This assumption usually does not hold because capillary condensation takes place and hides part of the initial area. The modification proposed lies on the observation that, when plotted as a function of either precoverage equilibrium pressure or adsorbed amount, the immersion energy of a nonporous solid in a wetting liquid drops down to a constant value once only 1.5-2 layers of water are preadsorbed. In these conditions, the area of the external water-vapor interface is closely comparable to that of the solid-water interface. Immersion of this system in water simply destroys the extemalliquid-water interface, whereas the energy of immersion is directly proportional to its area, after the following relationship:
dimmU = -A
( 'Ylv -
aYlV)
T aT
(12.18)
This equation is equivalent to Eqn (12.7), where ~sl - ~sv is replaced by ~lv· The method simply requires precovering the solid surface at a relative pressure corresponding to the plateau (a value slightly above 0.5 can be used safely
12.2
Immersion Calorimetry of Carbons into Pure Liquids
with water without any need for determining the whole graph of energy of immersion vs precoverage pressure). A surface area can then be determined without any assumption about the molecular cross section of the liquid. For a number of nonporous solids, the agreement was shown to be very good with the nitrogen-BET method. Nevertheless, one must keep in mind that the method requires the solid surface to be fully wetted by the liquid. In the case of carbons, a satisfactory wetting may require to use an alcane (like hexane) instead of water [13]. Furthermore, the method does not assess the surface area of micropores, which are filled during the precoverage process; in this case, it only assesses the "external" surface area, in a way similar to Sing's as method. We end this section with the characterization of microporou5 samples. An immersion experiment is a process where molecules initially in the bulk liquid are transferred to a solid-liquid interface. During this process, a number of liquidliquid bonds are transformed into liquid-solid bonds. The energetic balance, for the transfer of a molecule from the bulk liquid to a pore will depend very much on the relative size of the liquid molecule and the pore. If the pore size is such that only one molecule can penetrate, the enhancement of adsorbing potential will be of 2 and 3.68 for slit-shaped and cylindrical pores, respectively. Those calculations were performed by Everett and Powl [35] for the adsorption of one molecule interacting by only dispersive forces (Lennard-Jones type potential). Most interestingly, the above figures are very close to the ratio of the area covered by one molecule, in the corresponding pores, to its molecular cross section. These ratios are indeed 2 and 3.63, for slits and cylinders, respectively. It therefore looks appealing to extend this observation to any type of pore shape and to assume that, whatever it is, the immersion energy is simply proportional to the area accessible to the probe molecule [36] with a· coefficient which only depends on the solid-liquid system. This assumption can be checked, for pore sizes larger than a one molecular size, with help of the density functional theory (DFT). Because this is a thermodynamic calculation based on the minimization of the Grand Potential, the configurational energy is explicitly calculated with the DFT [37]. It is then possible to calculate the integral energy of adsorption. In the case of a porous system, the integral energy of adsorption up to completion of the pore filling can be related to the immersion energy by the following relationship:
where Llimmu is the immersion energy, Lladsu, the integral adsorption energy at saturation of the pore, na the amount adsorbed at saturation, and Llvapu the vaporization energy of the liquid. In Fig. 12.3, the integral adsorption energy per unit area calculated by DFT [38] is plotted as a function of the pore size, together with the corresponding immersion energy calculated by the preceding equation. This latter curve shows that the assumption ofproportionality between the surface area and the immersion energy holds for all pore size within ±10%. With a suitable nonporous standard of known surface area it is thus possible to determine the accessible surface area of any similar solid, even microporous.
288
Chapter 12 Porous Texture and Surface Characterization from Liquid-Solid Interactions
400 350 300
f
250
Integral energy of adsorption
"E' ;: 200 0)
CD
JJ
150 100
• • ••
•
Immersion energy
50
•
O+--------.------.---------r------.r----------,
o
5
10
15
20
25
Pore size (A)
Figure 12.3 Integral adsorption enthalpy and corresponding immersion enthalpy as a function of pore size, as calculated by DFT for the filling of a slit-shaped pore by a monoatomic fluid. (Adapted from [72].)
This method was shown to be well suited for microporous charcoals and immersed into organic liquids because of the absence of strong specific interactions [39, 40]. It is worth noting that for the smallest molecules used (benzene or methanol) the surface area provided by this method is, quite logically, higher than the nitrogen-BET equivalent surface area, since the BET method only takes into account one "side" of the molecule. Carrying out the same experiment for a set of liquids with different molecular sizes allows us to plot a graph of the accessible surface area as a function of pore width (Fig. 12.4). Assuming a pore shape, the next step is the derivation of a micropore width distribution, which can be compared to other approaches. One can indeed write d VCr)
= rdA(r) 2
where d V(r) and dA(r) are the pore volume and surface area in pores with sizes ranging between rand r + dr. For slit-shaped pores r is the aperture, whereas for cylinder r is the radius. The microporous volume between size a and b, is then given by
1
V
j 2
== -
b
rdA(r)
a
This integration can be performed on the curves in Fig. 12.4. The advantage of the method is that it gives a good assessment of the accessible pore volume by
12.3
Characterization of Carbons by Adsorption from Solution
1600 1400
0> .........
C\l
S
1200
C1
1000 C2
~
Q)
~
800
Q) ()
~
't:
600
:J
en
C3 400 - - . - - - - -.. C4
200 0 0.3
0.35
0.4
0.45
0.5
0.55
0.6
0.65
0.7
0.75
Pore width (nm)
Figure 12.4 Accessible surface area as a function ofpore width for a set ofactivated charcoals (activation increases from Cl to C4). The liquids used for immersion calorimetry are, in order of increasing size: benzene, methanol, isopropanol, cyclohexane, tertiary butanol, and a-pinene. (Adapted from [36].)
probing the solid at the temperature of interest, whereas the characterization of micropores by gas adsorption at 77 K may be limited by gas diffusion [4]. The immersion method was used for the study of zeolites [41] and it was recently extended to low-temperature immersion calorimetry, using nonspecific probes like liquid nitrogen, at 77 K, or liquid argon, at 87 K [42]. In some recent papers, immersion calorimetry is used in conjunction with gas adsorption (N2 or CO 2 ) to evidence gate effects that are observed when the pore entrance is partially blocked [43, 44].
12.3 CHARACTERIZATION OF CARBONS BY ADSORPTION FROM SOLUTION
Carbon materials are used in many industrial processes involving adsorption at a liquid-solid interface. Water purification by activated carbon, liquid chromatography, and stabilization of carbon black suspensions (inks, paints) are examples of such processes. The adsorption phenomena occurring at the solid-liquid interface are generally more complicated than those occurring at the solid-gas interface, simply because there is always competitive adsorption between at least two components. If the two components are miscible, the adsorption can be studied in the whole composition range (from 0 to 1 expressed in molar fraction).
290
Chapter 12 Porous Texture and Surface Characterization from Liquid-Solid Interactions
In the case of water solutions, the situation is extremely complex because water is itself a reactive solvent that is present under various forms (H 2 0, H+, or OH-) whose concentration depends on the pH. Moreover, adsorption is often studied in the presence of a salt, which also influences the adsorption process. Three species for water, two species for the salt in ionic form and one more for the solute then makes a minimum of six species involved in the adsorption process! As a consequence, a reliable study of adsorption from aqueous solution often requires to control or at least monitor pH, ionic strength, and temperature. Moreover, the concept of ionic strength may not be sufficient in the case where certain ions are specifically adsorbed. The reactivity of water with many surfaces, including carbon surfaces with polar groups, leads to the formation of a surface charge. The conditions of formation of the surface charge, its change with pH, ionic strength, and temperature were extensively studied in the case of nonporous minerals [15]. Another feature of adsorption from solution is the variety and complexity of molecules that may be involved in the processes. Indeed one can be interested either by a simple organic molecule, like benzene and its derivatives, or by much larger molecules like proteins, surfactants, or polymers, which bear many different chemical functions and may adopt a large number of conformations at the interface. For such molecules, a good knowledge of both the surface chemistry and the accessibility of porous materials are crucial to understand the adsorption phenomenon. In view of this complexity, here we shall focus our interest on aspects associated with the porosity of the solid. The first paragraph is about the basic concepts needed for such a kind of study. Examples are given in the second one. 12.3.1
Thermodynamics
In the field ofadsorption from solution, many discussions and reviews were published about the measurement of the adsorbed amount and the presentation of the corresponding data [14, 45-47]. Adsorption isotherms are the first step of any adsorption study. They are generally determined from the variation of macroscopic quantities which are rigorously measurable far away from the surface (e.g., the concentration of one species, the pressure, and the molar fraction). It is then only possible to compare two states: with or without adsorption. The adsorption data are derived from the difference between these two states, which means that only excess quantities are measurable. Adsorption results in the formation of a concentration profile near an interface. Simple representations are often used for this profile, but the real profile is an oscillating function of the distance from the surface [15, 16]. Without adsorption, the concentration should be constant up to the solid surface. Adsorption modifies the concentration profile of each component as well as the total concentration profile. It must be noted also that when the liquid is a pure component its concentration profile, i.e., its density, is also modified. Experimentally, the concentration can be measured at a large distance from the surface. The surface excess of component i is the
12.3
Characterization of Carbons by Adsorption from Solution
difference between the introduced amount ni and the amount calculated from the concentration measured far away from the surface c: and from a volume Vl,o which needs to be defined:
nC:1
= n. 1
c~1 VI,O
(12.19)
In the Gibbs representation, the volume V1,o is not limited by the solid adsorbent itself because the exact location of the adsorbing surface is actually unknown, so that this would introduce some uncertainty in the experimental data. Volume Vl,o is therefore limited by a fully theoretical surface (the Gibbs dividing surface, or GDS), which is precisely defined by the experimenter himsel£ although he usually tries to have it close to what he guesses to be the real adsorbing surface. What should not be forgotten when reporting liquid adsorption data (but which is rarely done) is therefore to state the exact way volume Vl,o was defined, in order to allow the reader to process the data with a different location of the GDS, which he may find more convenient to interpret the adsorption phenomenon. A way to avoid reporting this information is to eliminate Vl,o. This is possible after writing the preceding equation for each component or for the total amount of molecules, which leads to two possible ways to define and measure the surface excess: The relative suiface excess of 2 with respect to 1 (12.20)
the reduced suiface excess (12.21)
All surface excess amounts defined above usually refer to a unit mass or unit surface area (when available and when meaningful). The meaningfulness of the surface area requires being looked at thoroughly when porous solids like carbons are used for adsorbing large molecules from solution, because their surface areas were probably determined by gas adsorption of small molecules like nitrogen. By analogy with the characterization methods based on gas adsorption and on the shape of the isotherms, a classification of adsorption isotherms from liquid solution can be thought to be useful. The difficulties in establishing such a classification were underlined [9] .For dilute solutions Giles and Smith [48] proposed indeed 18 classes, Lyklema [15] simplified this down to 6, but we suggest retaining only 2 of them. Indeed, the shape of an adsorption isotherm from solution is the complex result of the balance between the solute-solute, solute-solvent, solute-surface, and surface-solvent interactions. Molecules do not only adsorb because they interact with the solid but also because the solvent may reject them. The surface is not itself a simple parameter because it is
292
Chapter 12 Porous Texture and Surface Characterization from Liquid-Solid Interactions
L
s
Figure 12.5 Two basic shapes of adsorption isotherms from dilute solutions. (Adapted from [9].)
generally heterogeneous. The presence of pores, of various crystalline faces or of different chemical sites influences the shape of the adsorption isotherm. It is only for a homogeneous surface that the relationship between the isotherm shape and the adsorption mechanism can be expected to be simple. One can then define two main shapes of isotherms (Fig. 12.5): L-type or S-type. The L-type, would follow the Langmuir model, which is site adsorption without any lateral interaction between the adsorbate molecules. The concavity of the curve, in normal scale, is always directed toward the concentration axis. The S-type would follow a more complex model in which lateral interactions between molecules are to be taken into account, using, e.g., the Bragg-Williams approximation [15]. A concavity of the adsorption isotherm directed toward the y-axis is a very strong indication of lateral interactions between molecules. If one looks at the IUPAC classification of gas adsorption isotherms [1], the same remark holds: this type of concavity is related with phenomena involving interactions between adsorbate molecules: capillary condensation, multilayer formation, 2-D phase changes, etc. Most experimental adsorption isotherms can be considered as a combination of these two "ideal" types. For heterogeneous surfaces, adsorption isotherms are often modeled as a combination of Land S adsorption isotherms corresponding to a distribution of patches [49, 50]. The many other shapes proposed in the preceding classifications [48] like stepwise, high affinity, or linear can be considered either as the combinations of S- and L-types or as a representation of the phenomenon for a limited range of concentration. For example, the highaffinity type is an extreme form of L-type. A linear adsorption isotherm (if it is not an artefact due to the penetration of the solute in the solid [15] may be the first portion of an L-type observed in the low concentration range. For the sake of characterization, only adsorption isotherms of simple shape may be used to provide safe interpretations. For example, to transform a surface excess amount into a surface area, a well-defined plateau is required, like in L-type isotherms for which a monolayer coverage can be assumed. If more complex shapes are obtained and if one wishes to extract from the data an
12.3
293
Characterization of Carbons by Adsorption from Solution
energetic distribution, one should then determine the adsorption enthalpy of the probe molecule. The sole adsorption isotherm usually does not allow, indeed, to estimate the role of surface heterogeneity and of the conformation changes as well as to discriminate between several mechanisms. There are two main ways to determine the adsorption enthalpy. One, called isosteric (because, for gas adsorption, it requires comparing two states with same amount adsorbed, i.e., same volume adsorbed), is the calculation of the differential adsorption enthalpy by using a set of two (or, better, three) adsorption isotherms at different temperatures. In dilute solution, the calculation of the isosteric enthalpy from adsorption isotherms at different temperatures is done by applying the following equation: . _
dadsh -
_
RT
2
(a In Xi) aT
(12.22) (T
nj
where the differentiation is performed while keeping constant all surface excess amounts. This condition makes it very difficult to apply this equation rigorously for liquid adsorption where, for instance, the surface charge varies with temperature. In the case of mixtures or concentrated solutions, activity coefficients have to be used. The second way to determine adsorption enthalpy is the direct measurement by microcalorimetry. Several papers are devoted to the analysis of the various ways to define liquid adsorption enthalpies and to measure them [51-55]. Experimentally, two types of calorimetric procedures can be distinguished on a thermodynamic basis: either the experiment is carried out in an open system or in a closed system. In the case of an open system, the main method consists in using a flowthrough setup. The sample is first equilibrated with the solvent, then with solutions of increasing concentration and, to end with, the desorption can be studied with a flow of pure solvent. Such an experiment mainly requires an equipment of chromatographic type, hence its name of "liquid frontal chromatography." The solid is placed in a column. Pumps are needed to inject solvent and solutions. It is possible to either prepare solutions in advance [56, 57] or to directly prepare various compositions by monitoring the flow rates of two pumps at constant total flow rate [51]. Downstream the column, a concentration detector (refractometer, UV, or IR spectrometer) allows recording the composition of the liquid as a function of time. Integration of the concentration profile vs time gives the reduced surface excess amount of the solute during one adsorption step. The limitations of this method are (i) that the sample grain size must fulfill some requirements (i.e., coarse enough to limit the pressure drop), and (ii) that the accuracy of the integration procedure depends on the long-term stability of the concentration recording. The main advantage is that all chemical potentials can be imposed throughout the experiment. This is important in pH-dependent experiments. Another interesting aspect is that such an experiment can be carried out in a microcalorimeter [56] giving
294
Chapter 12 Porous Texture and Surface Characterization from Liquid-Solid Interactions
both access to the surface excess amount and to the corresponding adsorption enthalpy. In the case of a closed system, the most common procedure is the immersion method [58]. It consists in immersing the solid in a liquid ofknown composition. After stirring and equilibration, the solid is separated from the liquid (centrifugation, filtration, or dialysis) and the final concentration is determined (UV, IR, refractometry or, still, Geiger counter in case radioactive tracers are used). Then the surface excess amount is calculated by using one of the equations derived in the preceding section. A direct analysis of the solid is also possible [59]. The immersion method is rewarding since it is simple (it does not need much equipment) though providing a good accuracy. It is also suited to follow the kinetics of adsorption. Another method using a closed system was devised by Nunn and Everett [60] with a flow-through equipment and a null procedure: the same solution is continuously circulated through the sample until equilibrium is attained, whereas a more concentrated solution is also injected to continuously restore the initial concentration. Since the concentration is continuously recorded, an independent kinetic experiment is not needed. For calorimetry, two different ways can be considered for closed systems experiments: • Immersion calorimetry of the dry solid in a solution (but this is not the safest way from a calorimetric viewpoint). • Direct determination of adsorption enthalpies (or more precisely displacement enthalpies as indicated earlier) by titration microcalorimetry, which is the main form of calorimetry used in adsorption from solution. For any method, care must be taken to define the reference state of the solute, which can be either the solution at equilibrium with the surface or the solution at infinite dilution state [4, 52]. An adsorption isotherm determined independently is needed to relate the calorimetric data with the surface excess amount. The most useful and convenient representation of calorimetric data shows the adsorption enthalpy as a function of surface concentration or coverage (or pore filling). Either integral or differential adsorption enthalpies can be determined. The integral enthalpy corresponds to the adsorption from zero coverage up to a given coverage. The differential enthalpy corresponds to the transfer of one mole of adsorbate from the bulk solution to the surface at a given coverage. In each case, the reference state can be either the equilibrium solution or the infinite dilution. The latter is suited when the properties of the solution are change much with concentration. This is the case with surfactant molecules, for instance. Strictly speaking, the above calorimetric experiments (either in closed or in open systems) provide "pseudodifferential" enthalpies of adsorption (rather than differential), because the actual experiments consist in discrete steps of surface concentration.
12.3
Characterization of Carbons by Adsorption from Solution
12.3.2
295
Applications
12.3.2.1
Surface area determination
Adsorption isotherms from solution have been used to determine the surface area of adsorbents for many years. Nevertheless, contrary to gas adsorption where nonspecific probes like argon or nitrogen can be used whatever the adsorbent, methods using adsorption from solution are generally specific for a class of material. For example, iodine or methylene blue are used for quick and convenient tests of adsorption capacity in the charcoal industry. A number of fundamental studies show how iodine [47, 61-63], p-nitrophenol [64, 65] salicylic acid [66], surfactants [67], or dyes [68] can be used for such applications. Nevertheless, most of these molecules present an affinity for the surface, which is highly dependent on the experimental conditions. For example, dyes and surfactants are very often electrically charged molecules and, because the adsorbing surface is also charged, the resulting adsorption isotherm depends on pH. A safe result cannot therefore be obtained from one experiment only. Also, the derivation of a surface area from a surface excess amount is based on the assumption that the average area per molecule is the same from one sample to the other. Because of these limitations we cannot specify a safe universal method to determine the surface area. Now, a particular feature ofadsorption from solution is the variety ofmolecules which can be used. Playing on their polarity or charge, it is then possible to define applications where the interest is not to determine the total surface area of the sample but, rather, to define the percentage of the surface, which can be considered as polar or nonpolar, hydrophilic or hydrophobic, acid or basic, etc. Groszek [69] extensively applied this approach, over 30 years, to the study of carbons Both calorimetry and adsorption isotherms may be used in such analysis. The influence of the surface charge may be very important on the adsorption from aqueous solution. In a recent review, Moreno-Castilla [70] gives examples of correlation ofthe adsorption data from solution (both isotherms and enthalpies of adsorption) and of the immersion energies with the amount of surface groups determined by an acid-base titration. These surface groups are generally directly related with the surface oxygen content of the carbon. 12.3.2.2
Pore size analysis
A way to proceed is to use probe molecules ofvarious sizes and to derive an accessible surface area from the amount adsorbed at the plateau of the adsorption isotherms [66]. Measurements with iodine showed that a method like the as plot, although originally devised for gas adsorption, could be extended to adsorption from solution. A set of adsorption isotherms, like those of Fig. 12.6, on various charcoals and on a nonporous reference sample was used to evaluate the method. The reference adsorption isotherm was normalized by dividing the amount adsorbed by the amount adsorbed at the plateau thus allowing a reference curve to be plotted as as vs equilibrium concentration. The amounts adsorbed on
296
Chapter 12 Porous Texture and Surface Characterization from Liquid-Solid Interactions
5
Microporous charcoal 4
4.5 C)
4
.........
(5
E
S "'0 CD
.0
0en
"'0 ctS
C :::J
0
E
3.5
Microporous charcoal 1
3 2.5 2 1.5
«
Nonporous carbon
0.5
0.00002
0.00004 0.00006 0.00008 0.0001 0.00012 Equilibrium concentration (mol/kg)
0.00014
0.00016
Figure 12.6 Adsorption isotherms of iodine on two microporous charcoals. (Adapted from [63].)
the charcoals were then plotted as a function of the as values. Plots similar to those for gas adsorption were obtained and allowed pore volumes and external surface areas to be calculated. The validity of the method was demonstrated by observing that when iodine completes the micropore filling its adsorption enthalpy becomes equal to that measured for the nonporous reference (Fig. 12.7). 80 70 (5
E
60
.........
~ ~ a.
50
co
E 40 CD
c: o
a
30
~
20
oen
10 O+------r----~-----.--------r--------.-----r-----.,
o
0.1
0.2
0.3
0.4 Coverage
0.5
0.6
0.7
Figure 12.7 Differential enthalpy of adsorption of iodine on two carbons (calorimetry). (Adapted from [63].)
References
297
For larger pores, say in the mesoporous range, much larger molecules are needed to characterize the pore size and the literature is scarce in this field. Polymers can be used (e.g., dextran) to evaluate the pore size of membranes. One then assesses a molar mass cutoff rather than a real pore size. The solute exclusion technique was also proposed to assess a pore size distribution [71]. It is well suited for wet porous materials [72].
REFERENCES 1. Rouquerol, J., Avnir, D., Fairbridge, C.W., et al. (1994). Recommendations for the characterization of porous solids. Pure Appl. Chern., 66(8), 1739-58. 2. Zettlemoyer, A.C., Chessick, J.J., and Hollabaugh, C.M. (1958). J. Phys. Chern., 62, 489-90. 3. Harkins, W.D. and Boyd, G.E. (1942). The binding energy between a crystalline solid and a liquid: the energy of adhesion and emersion. Energy of emersion of crystalline powders. II. J. Am. Chern. Soc., 64, 1195-204. 4. Rouquerol, F., Rouquerol, J., and Sing, K.S.W. (1999). Adsorption by Powders and Porous Solids: Principles, Methodology and Applications. Academic Press. 5. Sorensen, G.T. and Rouquerol, J. (eds) (2003). Sample Controlled Thermal Analysis: Origin, Goals, Multiple Forms, Applications and Future. Kluwer Academic Publishers. 6. Everett, D.H., Langdon, A.G., and Maher, P. (1984). Developments in immersion calorimetry- Design and testing of an improved sel-breaking technique. J. Chern. Thermodynamics, 16, 981-92. 7. Partyka, S., Rouquerol, F., and Rouquerol, J. (1979). Calorimetric determination of surface areas: Possibilities of a modified Harkins and Jura procedure. J. Colloid Interface Sci., 68(1), 21-31. 8. Gomez, F., Denoyel, R., and Rouquerol, J. (2000). Determining the contact angle of a nonwetting liquid in pores by liquid intrusion calorimetry. Langmuir, 16, 4374-9. 9. Denoyel, R. and Rouquerol, F. (2002). Adsorption from the liquid phase. Handbook of Porous Solids. Wiley-VCH, Chapter 2.6. 10. Douillard,J.M. and Zajac,]. (2006). Contact Angle Determination from Heat of 1mmersion and Heat of Wetting. Encyclopedia of Surface and Colloid Science. Marcel Dekker. 11. Chessick, J.J. and Zettlemoyer, A.C. (1959). Immersional heats and the nature of solid surfaces. Adv. Catal., 11, 263-99. 12. Zettlemoyer, A.C. (1965). Immersional wetting of solid surfaces. Ind. Eng. Chern., 57, 27-36. 13. Harkins, W.D. and Jura, G. (1944). Surfaces of solids. XII. An absolute method for the determination of the area of a finely divided crystalline solid. J. Am. Chern. Soc., 66, 1362-6. 14. Everett, D .H. (1972). Manual ofsymbols and terminology for physicochemical quantities and units. Appendix 2, Part 1. Pure Appl. Chern., 31(4), 579-638. 15. Lyklema, J. (1995). Fundamentals of Interface and Colloid Science. I Fundamentals. II. SolidLiquid Interfaces. London: Academic Press. 16. Israelachvili, J.N. (1992). Intermolecular and Surface Forces. Academic Press. 17. Medout-Madere, V. (2000). A simple experimental way of measuring the Hamaker constant All of divided solids by immersion calorimetry in apolar liquids. J. Colloid Interface Sci., 228, 434-7.
298
Chapter 12 Porous Texture and Surface Characterization from Liquid-Solid Interactions
18. Zettlemoyer, A.C., Chessick, J.J., and Hollabaugh, C.M. (1958). Estimation of the surface polarity ofsolids from heat of wetting measurements.]. Phys. Chem., 62, 489-90. 19. Morimoto, T. and Suda, Y. (1985). Heat of immersion of zinc oxide in organic liquids. 1. Effect of surface hydroxyls on the electrostatic field strength. Langmuir, 1, 239-43. 20. Young, G.J., Chessick, J.J., Healey, F.H., and Zettlemoyer, A.C. (1954). Thermodynamics of the adsorption ofwater on Graphon from heats ofimmersion and adsorption data.]. Phys. Chem., 58, 313-15. 21. Healey, F.H., Yu, Y.F., and Chessick, J.J. (1955). The detection of hydrophilic heterogeneities on a carbon surface.]. Phys. Chem., 59, 399-402. 22. Lopez-Ramon, M.V., Stoeckli, F., Moreno-Castilla, C., and Carrasco-Marin, F. (1999). On the characterization of acidic and basic surface sites on carbons by various techniques. Carbon, 37, 1215-21. 23. Robert, L. and Brusset, H. (1965). Heat of immersion of carbon products. Fuel, 44, 309-16. 24. Rodriguez-Reinoso, F. and Molina-Sabio, M. (1998). Textural and chemical characterization of microporous carbons. Adv. Colloid Inteiface Sci., 76-77, 271-94. 25. Lopez-Ramon, M.V., Stoeckli, F., Moreno-Castilla, C., and Carrasco-Marin, F. (2000). Specific and non-specific interactions of water molecules with carbon surfaces from immersion calorimetry. Carbon, 38, 825-9. 26. Stoeckli, F., Moreno-Castilla, C., Carrasco-Marin, F., and Lopez-Ramon, M.V. (2001). Distribution ofsurface oxygen complexes on activated carbons from immersion calorimetry, titration and temperature-programmed desorption techniques, Carbon, 39(14), 2235-7. 27. Szymaski, G.S., Biniak, S., and Rychlicki, G. (2002). Carbon surface polarity from immersion calorimetry. Fuel Process. Technol., 79(3), 217-23. 28. Bradley, R.H., Daley, R., and Le Gof£ F. (2002). Polar and dispersion interactions at carbon surfaces: further development of the XPS-based model. Carbon, 40(8), 1173-9. 29. Gonzalez-Martin, M.L., Gonzalez-Garcia, C.M., Gonzalez, J.F., et al. (2002). Thermodynamic characterization of a regenerated activated carbon surface. Appl. Suif. Sci., 191, 166-70. 30. Briant, J. and Cuiec, L. (1972). Comptes-Rendus du 4eme Colloque ARTEP, RueilMalmaison, 7-9 ]uin 1971. Paris: Ed. Technip. 31. Douillard, J.M., Zoungrana, T., and Partyka, S. (1995). Surface Gibbs free energy of minerals: some values.]. Petrol. Sci. Eng., 14,51-7. 32. Medout-Marere, V., Partyka, S., Dutartre, R., et al. (2003). Surface heterogeneity of passively oxidized silicon carbide particles: vapor adsorption isotherms.]. Colloid Inteiface Sci., 262, 309-20. 33. Spagnolo, D.A., Maham, Y., and Chuang, K.T. (1996). Calculation of contact angle for hydrophobic powders using heat of immersion data.]. Phys. Chem., 100, 6626-30. 34. Neumann, A.W. (1974). Contact angles and their temperature dependence: thermodynamic status, measurement, interpretation and application. Adv. Colloid Inteiface Sci., 4, 105-91. 35. Everett, D.H. and Powl, J.C. (1976). Adsorption in slit-like and cylindrical micropores in Henrys law region - Model for microporosity of carbons.]. Chem. Soc. Faraday Trans. I, 72, 619-36. 36. Denoyel, R., Fernandez-Colinas, J., Grillet, Y., and Rouquerol, J. (1993). Assessment of the surface area and microporosity of activated charcoals from immersion calorimetry and nitrogen adsorption data. Langmuir, 9, 515-18. 37. Olivier, J.P. (2000). Comparison of the experimental isosteric heat of adsorption on mesoporous silica with density functional theory calculations. Stud. Suif. Sci. Catal., 128. 81-7.
References
299
38. Denoyel, R., Beurroies, I., and Vincent, D. (2000). Microcalorimetric methods for studying vapour adsorption and wetting of powders. J. Thermal Anal. Calorimetry, 70, 483-92. 39. Neugebauer, N.N. (1999). PhD Dissertation, Leipzig University, Germany. 40. Rodriguez-Reinoso, F., Molina Sabio, M., and Gonzalez, M.T. (1997). Effect of oxygen surface groups on the immersion enthalpy of activated carbons in liquids of different polarity. Langmuir, 13, 2354-8. 41. Silvestre-Albero, J., Gomez de Salazar, C., Sepulveda-Escribano, A., and RodriguezReinoso, F. (2001). Characterization of microporous solids by immersion calorimetry. Colloids Suif. A: Physicochem. Eng. Aspects, 187-188, 151-65. 42. Rouquerol, J., Llewellyn, P., Navarette, R., et al. (2002). Assessing microporosity by immersion microcalorimetry into liquid nitrogen or liquid argon. Stud. Sutj. Sci. Catal., 144, 171-6. 43. Cagnon, B., Py, X., Guillot, A., and Stoeckli, F. (2003). The effect of the carbonization/activation procedure on the microporous texture of the subsequent chars and active carbons. Microporous Mesoporous Mater., 57, 273-82. 44. Stoeckli, F., Slasli, A., Hugi-Cleary, D., and Guillot, A. (2002). The characterization of microporosity in carbons with molecular sieve effects. Microporous Mesoporous Mater., 51(3), 197-202. 45. Defay, R. and Prigogine, 1. (1951). Tension Supeificielle et Adsorption. Liege-Paris: DesoerDunod. 46. Schay G. (1970). In Proceedings of the International Symposium on Suiface Area Determination (D.H. Everett and R.H. Otterwill, eds). London: Butterworth, p. 273. 47. Kipling, J.J. (1965). Adsorption from Solution of Non-electrolytes. London: Academic Press. 48. Giles, C.H. and Smith, D. (1974). A general treatment and classification of the solute adsorption isotherm. 1. Theoretical]. Colloid Inteiface Sci., 47, 755-65. 49. Cases, J .M. (1979). Tensio-active adsorption at the solid-liquid interface - Thermodynamics and influence of adsorbant heterogeneity. Bull. Mineralogie, 102, 684-707. 50. Cases, J.M. and Villieras, F. (1992). Thermodynamic model of ionic and non-ionic surfactants adsorption-abstraction on heterogeneous surfaces. Langmuir, 8, 1251-64. 51. Johnson, 1., Denoyel, R., Everett, D .H., and Rouquerol, J. (1990). Adsorption at the liquid/graphite interface: Comparison of enthalpy data obtained from three different methods. Colloids Suif-, 49, 133-48. 52. Denoyel, R., Rouquerol, F., and Rouquerol, J. (1990). Thermodynamics of adsorption from solution: Experimental and formal assessment of the enthalpies of displacement. ]. Colloid Inteiface Sci., 136, 375-84. 53. Kiraly, Z. and Dekany, 1. (1989). Thermodynamics of multilayer adsorption of aqueous butanol solution onto printex and graphitized printex carbon-blacks.]. Chem. Soc. Faraday Trans. I, 85, 3373-83. 54. Kiraly, Z., Dekany, 1., and Nagy, L.G. (1993). Thermodynamic formulation of adsorption phenomena at the solid/solution interface: A practical approach. Colloids Sutj. A, 71,287-92. 55. Woodbury, G.W. and Noll, L.A. (1987). Heats of adsorption from flow calorimetry: Relationships between heats measured by different methods. Colloids Sutj., 28, 233-45. 56. Denoyel, R., Rouquerol, F., and Rouquerol, J. (1983). Interest and requirements of liquid-flow microcalorimetry in the study of adsorption from solution in the scope of tertiary oil recovery. In Adsorption from Solution (C. Rochester and A.L. Smith, eds.). Academic Press, pp. 225-34. 57. Kiraly, Z. and Findenegg, G.H. (1998). Calorimetric evidence of the formation of halfcylindrical aggregates of a cationic surfactant at the graphite/water interface.]. Phys. Chem. B., 102,1203-11.
300
Chapter 12 Porous Texture and Surface Characterization from Liquid-Solid Interactions
58. Everett, D .H. (1986). Reporting data on adsorption from solution at the solidisolution interface. Pure Appl. Chem., 58(7), 967-84. 59. Nunn, C. Schlechter, R.S and Wade, W.H. (1981). A direct method for measuring adsorption from solution onto solids. J. Colloid Interface Sci., 80, 598-605. 60. Nunn, C. and Everett, D.H. (1983). A note on the determination of adsorption from solution. J. Chem. Soc. Faraday Trans 1., 79, 2953-4. 61. Puri, B.R. and Bansal, R.C. (1965). Iodine adsorption method for measuring surface area of carbon blacks. Carbon, 3, 227-30. 62. Molina-Sabio, M., Salinas-Martinez de Lecea, C., et al. (1985). A comparison of different tests to evaluate the apparent surface area of activated carbons. Carbon, 23, 91-6. 63. Fernandez-Colinas, J., Denoyel, R., and Rouquerol, J. (1989). Adsorption of iodine from aqueous solutions onto activated carbons: correlations with nitrogen adsorption at 77 K. Adsorp. Sci. Technol., 6, 18-26. 64. Giles, C.H. and Nakhwa, S.N. (1962). Adsorption XVI The measurement of specific surface areas of finely divided solids by solution adsorption. J. Appl. Chem., 12, 266-73. 65. Lopez-Gonzalez, J., de, D., Valenzuela-Calahorro, C., et al. (1988). Adsorption of p-nitrophenol by active carbons prepared from olive wood. An. Quim., 84B, 47-51. 66. Femandez-Colinas, J., Denoyel, R., and Rouquerol, J. (1991). Characterization of activated charcoals by adsorption from solution. Stud. Surf. Sci. Catal., 62, 399-408. 67. Somasundaran, P. and Fuerstenau, D.W. (1966). Mechanisms of alkyl sulfonate adsorption at the alumina-water interface. J. Phys. Chem., 70, 90-6. 68. Giles, C.H., D'Silva, A.P., and Stridevi, A. (1969). In Proceedings of the International Symposium on Surface Area Determination (D.H. Everett and R.H. Ottewill eds). London: Butterworths, pp. 317-23. 69. Groszek, AJ. (1998). Flow adsorption microcalorimetry. Thermochim. Acta, 313, 133-43. 70. Moreno-Castilla, C. (2004). Adsorption of organic molecules from aqueous solutions on carbon materials. Carbon, 42, 83-94. 71. Lin, J.K., Ladish, M.R., Patterson, J.A., and Noller, C.H. (1987). Determining poresize distribution in wet cellulose by measuring solute exclusion using a differential refractometer. Biotechnol. Bioeng., 29, 976-81. 72. Denoyel, R. (2004). Adsorption of organic molecules in nanoporous adsorbents from aqueous solution. In Nanoporous Materials: Science and Engineering. (G.Q. Lu and S. Zhao, eds.). World Scientific, Ch. 23, pp. 727-55.
SURFACE CHEMICAL CHARACTERIZATION OF
CARBONS FROM ADSORPTION STUDIES Hans-Peter Boehm Department
of Chemistry and Biochemistry,
University
of Munich,
Germany
Contents 13.1 Introduction 13.2 Hydrophilic Carbon Surfaces 13.3 Surface Oxides of Carbon 13.4 Amphoteric Character of Carbons 13.5 Electrokinetic Phenomena 13.6 Effects on the Adsorption of Inorganic ions References
301 302 30 4 30 8 31 8 321
32 3
13.1 INTRODUCTION
The adsorption behavior of carbons is affected to a considerable extent by the chemical state of their surfaces, which is also of great practical importance in many other applications of carbon materials such as for catalysts and catalyst supports, and carbon-polymer composites. In the surface of carbon materials, the regular network of covalent C-C bonds is broken, and reactive sites result as a consequence. Usually "free valences," also called "dangling bonds," are saturated with foreign elements, in first line hydrogen and oxygen. In the case of carbon structures derived from the graphite lattice, the surface is inhomogeneous and is constituted to variable fractions of basal faces, i.e., honeycomb-like graphene layers, and of the edges of the graphene layers. While the basal faces are quite inert, the edge sites are reactive and can chemisorb other elements such as hydrogen, oxygen, nitrogen species, and halogens. In contrast, the surface of Adsorption by Carbons ISBN: 978-0-08-044464-2
© 2008 Elsevier Ltd.
All rights reserved.
301
302
Chapter 13 Surface Chemical Characterization of Carbons from Adsorption Studies
diamond is much more homogeneous and has a comparatively simpler chemical behavior. One would expect that many free radical sites (dangling bonds) exist on an atomically clean surface, but the number of free radicals determined by electron spin resonance (ESR) measurements on carbons is much smaller than corresponds to the estimated number of edge sites, and a part of them may also be located at vacancy sites within the graphene layers [1-3]. One reason might be that atomically clean surfaces of solids are frequently reconstructed, leading to new electronic states that can accommodate electron pairs. Also the localization of 11" electrons at free radical edge sites with formation of carbene-like structures (in-plane sigma pairs) has been suggested (see p. 229 in Ref [1] and Fig. 3 in Ref [4]). Measurement ofadsorption phenomena by chemical means require adsorbents that have a relatively high surface area, preferably in excess of 20-50 m 2 Ig, to provide sufficient sensitivity. Such carbons are, e.g., activated carbons, carbon blacks, graphite wear dust, and carbon nanotubes. Physical measurements, such as by X-ray photoelectron spectroscopy (XPS) , Auger electron spectroscopy (AES) , electron energy loss spectroscopy (EELS), Fourier transform infrared (FTIR), and special Raman spectroscopies, can be done with materials of much lower surface area.
13.2 HYDROPHILIC (ARBON SURFACES The basal faces and chemically"clean" edge faces, as well as those saturated with chemisorbed hydrogen, are hydrophobic, whereas surfaces with oxygencontaining surface groups are hydrophilic. Clean carbon surfaces with a surface roughness in the 40-50 nm range are "superhydrophobic" (Lotus effect), i.e., they have a contact angle with water of > 150 [5]. The hydrophilicity of a carbon surface determines its adsorption behavior toward water vapor. Hydrophobic surfaces show type III adsorption isotherms (type V in the case of porous carbons) [6]. Very little water is adsorbed at low relative pressures plpo because adsorption occurs only by dispersion forces, no hydrogen bonds can be formed. Water behaves similar to a gas of low molecular mass, such as neon, with a correspondingly low boiling point. At room temperature it would be supercritical. Only when some water molecules are adsorbed, the following ones can form hydrogen bonds, and larger water clusters are formed on the surface. At higher relative pressures, the nature of the adsorbed water will gradually change to that of normal liquid, hydrogen-bonded water, and the adsorption isotherms rise steeply at plpo values above 0.5. In the case of microporous carbons, adsorption is promoted by the higher adsorption potential in narrow pores, and the adsorption isotherms begin to rise steeply at much lower plpo values (type V isotherms). Consequently, one might expect that a superposition of type II and type III isotherms might occur if there existed only 0
13.2
303
Hydrophilic Carbon Surfaces
Adsorbed water (mg/g) 4.0
3.0
2.0
1.0
0.5
1.0
p/Po Water vapor adsorption isotherms on diamond powder (20m2 /g). (1) Treated with H 2 at 800°C (measured at 17.8 °C); (2) outgassed in vacuo at 900°C (18.3 °C); (3) oxidized with 02 at 420°C (19.8 °C). (Reprinted from Re£ [8] with permission from Elsevier.)
Figure
13.1
a small concentration of hydrophilic adsorption sites (chemisorbed oxygen) on an otherwise hydrophobic nonporous carbon. Figure 13.1 shows water vapor adsorption isotherms on a clean, a hydrogenated and an oxidized diamond surface. Clearly, water adsorption is promoted by surface oxygen complexes whilst the hydrogenated surface is the most hydrophobic one. If there is a superposition of type II and type III isotherms, it should be possible to estimate the concentration of hydrophilic sites by application of the Brunauer-Emmett-Teller (BET) adsorption equation [7] at relatively low plpo values. The BET equation is based on the assumption that the heat of adsorption is significantly higher in the first adsorbed layer than in the following adlayers where it is practically equal to the heat of liquefaction [7]. In our case, the first adlayer corresponds to the adsorption of one water molecule on each hydrophilic adsorption site. The heat of adsorption is significantly higher than in the following adsorption. Indeed, an excellent correlation of hydrophilic sites determined by application of the BET equation with active hydrogen (of hydroxyl groups) determined by independent methods was observed on the surface of diamond powder [8] and also on oxidized SiC [9], as summarized in Table 13.1 this method is not applicable, however, when there are higher concentrations of hydrophilic adsorption sites.
Chapter 13 Surface Chemical Characterization of Carbons from Adsorption Studies
304
Table 13.1 Hydrophilic adsorption centers on the surface of diamond, SiC, and pyrogenic silica (Aerosil) (Data from Refs [8, 9].)
Diamond, outgassed at 900°C (20 m 2 / g) Diamond, H 2 at 800°C (20 m 2 / g) Diamond, oxidized with O 2 at 420°C (20m2 /g) SiC, oxidized with air at 20°C (9.5m 2 /g) Si0 2 (Aerosil) a
b C
19
n.d.
n.d.
7
n.d.
n.d.
62
66
53
54
2160
2210
From weight increase on isotope exchange with D 2 0 (-OH to -OD). By reaction with CH 3 MgI (volumetric determination of evolved CH 4 ). By titration with NaOH of weakly acidic surface groups.
13.3
SURFACE OXIDES OF (ARBON
The most intensively studied surface complexes of carbons are those with oxygen. Such surface oxides can be produced by treatment with gaseous oxidants such as dioxygen (or air), ozone, oxygen plasma, or NO x ' Dioxygen molecules react only with carbon atoms at the edges the graphene layers or at defects, e.g., vacancies, within the planes [10-12]. The surface layers are, however, attacked by free radicals, atomic oxygen, and compounds that easily produce atomic oxygen by decomposition such as ozone. For instance, large, millimeter-sized flakes of well-crystallized graphite were converted to an evil-smelling sludge on prolonged exposition to UV-irradiated CC1 4 [13] Cl, and CC1 3 radicals are formed by photolysis.
13.3.1 Generation of Surface Oxides Reaction temperatures of 250-400 °C are usually taken for oxidation with dioxygen or air. Significant quantities of surface oxides of mostly acidic character are produced in a few hours. The required temperatures are the lower, the smaller the particle size of the carbon is. Clean surfaces of turbostratic carbons of high surface area such as activated carbons and carbon blacks will also be oxidized at room temperature. When the carbons are freed from surface complexes by heating to 900-1000°C in vacuo or under an inert gas, they will adsorb some
13.3 Surface Oxides of Carbon
305
oxygen on exposure to air at room temperature. The reaction is quite fast in the beginning, but slows down gradually [14] (see Section 13.4.2). Much more oxygen is bound on the surface in a slow reaction with moist air. This phenomenon, called "aging," was first described by Puri [15]. The presence of water vapor accelerates the reaction significantly [16-18]. The aging process takes several months at room temperature to become easily measurable. It can be followed easily within a few weeks when the reaction occurs at mildly raised temperatures [19,20]. as shown in Fig. 13.2. An activated carbon (Norit) and a furnace black (Corax 3) were oxidized in air of 85% relative humidity at 60°C or under ambient air of varying humidity at 110°C. Sodium hydroxide uptake was used as a measure of aging since acidic groups are formed in the reaction (see below). The figure shows clearly, that in the case of the activated carbon the surface oxidation occurred faster and to a higher extent at the lower temperature at higher relative humidity than at the higher temperature at a much lower relative humidity [19]. Aging is drastically increased when catalytically active metals, e.g., palladium, are deposited on the surface (Fig. 13.2). This aging process causes changes in the properties of carbon materials. The surface becomes more and more hydrophilic, and the adsorption capacity of activated carbons for noxious gases or methyl iodide is greatly reduced [16, 20]. With porous carbons, the surface oxidation begins at the outer surface of the particles, but progresses very slowly into their interior due to very slow diffusion of oxygen in narrow pores [21]. In consequence, the exterior and interior surfaces of activated carbons can differ significantly in their adsorption properties. Aging can
(a)
(b)
NaOH uptake
NaOH uptake
(Jlmol/g)
(Jlmol/g)
400
250
.
--~.=-------
300 150
.............
././
200 100
ro
50
20
40
60 days
- - Time
_-......_._ ...
_
-.
......... ....
7, trivalent ions tend to hydrolyze already
322
Chapter 13 Surface Chemical Characterization of Carbons from Adsorption Studies
at pH values in the acidic range. However, in acidic solutions H 30+ ions will compete with other cations, and weakly acidic groups will not be dissociated. For instance, hydrated aluminum ions will form hydroxo-pentaquo complexes: (13.9) Further hydrolysis leads to binuclear and oligonuclear complex ions. The situation is similar with [Fe(HzO)6]3+. Only [Cr(HzO)6]3+ is kinetically stable at neutral pH. In the so-called basic aluminum chloride solutions at pH 4-8, the cation [Al 13 0 4(OH)Z4(H zO)lZ]7+ is formed [81]. It has a spherical Keggin-like structure with a central tetrahedrally coordinated Al 3+ ion, analogous to the structure of the phosphomolybdate anion [PMo 1Z 0 40 ]3- . Larger complex ions, in particular polynuclear ones, can close the entrances to narrow micropores and reduce the surface area available for adsorption. On the other hand, the adsorption of hydrated cations, and especially of larger, polynuclear cations, will be promoted by formation of hydrogen bonds to oxygen or hydroxyl surface functions close to the negatively charged surface site. In consequence, no simple relationship for binding by a specific surface group can be established. Carbonyl or hydroxyl groups on the carbon surface can also replace water molecules of the hydration shell of the cations, giving rise to bidentate adsorption sites. The matter becomes still more complicated with heavy-metal cations that have a "soft", easily polarizable electron shell that can also interact with the 1T electron systems of the graphene layers. Quite a large number of publications have appeared dealing with the adsorption of hydrated or otherwise coordinated metal cations for the purification of contaminated waste waters. However, the experimental conditions can not be compared with those described in the preceding sections since usually very dilute solutions are used (often in the range of 10- 5 - 10- 4 M) to simulate realistic conditions (see also Chapter 25 and Refs [4, 82]). Although no direct correlation with the number of specific surface groups can be found, in general the adsorption of metal cations increases with the concentration of acidic surface functions [24]. Infrared spectroscopic experiments showed that free carboxyl groups absorbing at 1717 cm-1 are converted to the ionized carboxylate form (1576cm- 1) on adsorption ofCd z+ ions [24]. The adsorption capacity for transition metal ions can be further increased by introducing nitrogen surface groups, e.g., by treatment with ammonia at high temperatures [83]. Increased adsorption, compared to the activated carbons in oxidized form, was observed with Cd 2 +, Ni z+, and Cu z+ ions. The authors suggest that pyridine-type nitrogen on the edge of the carbon layers is responsible. In particular pairs of nitrogen atoms in a situation analogous to that in 1,10-phenanthroline would allow very stable bidentate coordination. As mentioned in Section 13.3.1, precious metal complex ions can be reduced by the carbon to the metal [28-31]. An interesting phenomenon was observed with [Au(CN)z]- solutions that playa role in the leaching of gold ores. Clearly, such ions are bound by the basic sites of the surface. XPS studies of the adsorbed
References
gold species showed relatively sharp Au 4f peaks. The Au 4f7 / 2 binding energy of 91.5 eV, however, was in between that of elementary gold (90.8 eV) and that ofK[Au(CN)2J (93.8eV), see Ref. [84] or Section 6.8.1.4 in Ref. [82]. One can conclude that the gold is not reduced to the elementary state, but to a gold cluster compound [Aun(CN)m]X- in which gold has formally a fractional oxidation number smaller than 1.
REFERENCES 1. Leon y Leon, C.A. and Radovic, L.R. (1994). Interfacial chemistry and electrochemistry of carbon surfaces. In Chemistry and Physics of Carbon, Vol. 24 (P .A. Thrower, ed.). Marcel Dekker, pp. 213-310. 2. Mrozowski, S. (1971). Electronic properties and band model of carbons. Carbon, 9,97-109. 3. Mrozowski, S. (1988). ESR studies of carbonization and coalification processes. Part I. Carbonaceous compounds. Carbon, 26, 521-9. 4. Radovic, L.R., Moreno-Castilla, C., and Rivera-Utrilla,J. (2000) Carbon materials as adsorbents in aqueous solutions. In Chemistry and Physics of Carbon, VoL 27 (P .A. Thrower, ed.). Marcel Dekker, pp. 227-405. 5. Feng, L. and Yang, Z.L. (2003). Superhydrophobicity of nanostructered carbon films in a wide range of pH values. Angew. Chern., Int. Ed. Engl., 42, 4217-20. 6. Gregg, S.J. and Sing, K.S.W. (1982). Adsorption, Suiface Area and Porosity, 2nd edn. Academic Press, Chapter 5. 7. Brunauer, S., Emmett, P.H., and Teller, E. (1938). Adsorption of gases in multimolecular layers. J. Am. Chern. Soc., 60, 309-19. 8. Sappok, R. and Boehm, H.-P. (1968). Surface chemistry of diamond - II. Formation, properties and structure of the surface oxides. Carbon, 6, 573-88. 9. Sappok, R. and Boehm, H.-P. (1969). Gravimetric determination of active hydrogen on the surface of solids (in German). Z. Anolg. AUg. Chern., 365, 152-6. 10. Hennig, G.R. (1966). Electron microscopy of reactivity changes near lattice defects in graphite. In Chemistry and Physics of Carbon, Vol. 2 (P.L. Walker, ]r, ed.). Marcel Dekker, pp. 1-49. 11. Thomas,].M. (1965). Microscopic studies of graphite oxidation. In Chemistry and Physics of Carbon, Vol. 1 (P.L. Walker,]r, ed.). Marcel Dekker, pp. 121-202. 12. Yang, R.T. (1984). Etch-decoration electron microscopy studies of the gas-carbon reaction. In Chemistry and Physics of Carbon, Vol. 19 (P.A. Thrower, ed.). Marcel Dekker, pp. 163-210. 13. Schlagl, R. and Boehm, H.-P. (1988). Photochemical intercalation in graphite. Synth. Metals, 23, 407-13. 14. Voll, M. and Boehm, H.-P. (1970). Basic surface oxides on carbon - II. Stoichiometry and kinetics of the formation reaction; thermal decomposition (in German). Carbon, 8, 741-52. 15. Puri, B.R. (1962). Surface oxidation of charcoals at ordinary temperatures. In Proceedings 5th Biennial Conference on Carbon, Pennsylvania State University, University Park, PA, 1961. Pergamon Press, pp. 165-70.
Chapter 13 Surface Chemical Characterization of Carbons from Adsorption Studies
16. Billinge, B.H.M., Docherty, J.B., and Bevan, M.J. (1984). The desorption of chemisorbed oxygen from activated carbons and its relationship to ageing and methyl iodide retention efficiency. Carbon, 22, 83-9. 17. Deitz, V.A. (1987). Interaction of radioactive iodine gaseous species with nucleargrade activated carbons. Carbon, 25, 31-8. 18. Adams, L.B., Hull, C.R., Holmes, R.J., and Newton, R.A. (1988). An examination of how exposure to humid air can result in changes in the adsorption properties of activated carbons. Carbon, 26, 451-9. 19. Kuretzky, T. and Boehm, H.-P. (1994). Effect of water vapor and of platinum metals on the aging of carbon at moderate temperatures. In Extended Abstracts and Programme, Carbon'94, International Carbon Conference, Granada, Spain. University of Granada Press, pp. 262-3. 20. Billinge, B.H.M. and Evans, M.G. (1984). The growth ofsurface oxygen complexes on the surface of activated carbon exposed to moist air and their effect on methyl iodide-131 retention.]: Chimie Physique, Physico-Chimie Biologique, 81, 779-84. 21. Menendez,J.A., Illin-Gomez, M.J., Leon y Leon, C.A., and Radovic, L.R. (1995). On the difference between the isoelectric point and the point of zero charge of carbon. Carbon, 33, 1655-7. 22. Verma, S.K. and Walker, P.L., Jr (1992). Carbon molecular sieves with stable hydrophobic surfaces. Carbon, 30, 837-44. 23. Menendez,J.A., Phillips,J. Xia, B., and Radovic, L.R. (1996). On the modification and characterization of chemical surface properties ofactivated carbon: In the search of carbons with stable basic properties. Langmuir, 12, 4404-10. 24. Jia, Y.F. and Thomas, K.M. (2000). Adsorption of cadmium ions on oxygen surface sites in activated carbon. Langmuir, 16, 1114-22. 25. Strelko, V., Malik, D.J., and Streat, M. (2002). Characterisation of the surface of oxidised carbon adsorbents. Carbon, 40, 95-104. 26. Moreno-Catilla, C., Carrasco-Marin, F., and Mueden, A. (1997). The creation of acidic carbon surfaces by treatment with (NH4)2S20s. Carbon, 35, 1619-26. 27. Sepulveda-Escribano, A., Coloma, F., and Rodriguez-Reinoso, F. (1998). Platinum catalyst supported on carbon blacks with different surface chemical properties. Appl. Catal. A (General), 173,247-57. 28. Puri, B., Singh, S., and Mahajan, O.P. (1965). Interaction of charcoal with ammoniacal and aqueous silver nitrate. Indian]. Chem., 3, 54-7. 29. Suh, D.J., Park, T.-J., and Ihm, S.-K. (1992). Characterization of carbon-supported palladium catalysts for liquid-phase hydrogenation of nitroaromatics. Industr. Eng. Chem. Res., 31, 1849-56. 30. Fu, R., Zeng, H., and Lu, Y. (1994). Studies on the mechanism of the reaction of activated carbon fibers with oxidants. Carbon, 32, 593-8. 31. Gurrath, M., Kuretzky, T., Boehm, H.-P., et al. (2000). Palladium catalysts on activated carbon supports: Influence of reduction temperature, origin of the support and pretreatments of the carbon surface. Carbon, 38, 1241-55. 32. Vue, Z.R., Jiang, W., Wang, L., et al. (1999). Surface characterization of electrochemically oxidized carbon fibers. Carbon, 37. 1785-96. 33. Jannakoudakis, A.D., Jannakoudakis, P.D., Theodoridou, E., and Besenhard, J.O. (1990). Electrochemical oxidation of carbon fibers in aqueous solutions and analysis of the surface oxides.]. Appl. Electrochem., 20, 619-24. 34. Donnet, J.-B. and Bansal, R.C. (1990). Carbon Fibers, 2nd edn. Marcel Dekker.
References
325
35. Kozlowski, C. and Sherwood, P.M.A. (1987). X-ray photoelectron spectroscopic studies of carbon fiber surfaces. VIII. - A comparison of type I and type II fibers and their interaction with thin resin films. Carbon, 25, 751-60. 36. Barton, S.S. and Evans, M.J.B., Halliop, E. and MacDonald, J.A.F. (1997). Anodic oxidation of porous carbon. Langmuir, 13, 1332-6. 37. Boehm, H.-P. (1994). Some aspects of the surface chemistry of carbon blacks and other carbons. Carbon, 32, 759-69. 38. Boehm, H.-P. (2002). Surface oxides on carbon and their analysis: a critical assessment. Carbon, 40, 145-9. 39. Kruyt, H.R. and de Kadt, G.S. (1929). The electric charge of colloidal carbon (in German). Kolloid-Z., 47, 44. 40. Kolthof"L I.M. (1932). Properties of active charcoal reactivated in oxygen at 400°C. J. Am. Chern. Soc., 54, 4473-80. 41. Newman, M.S. and Muth, C.W. (1951). Normal and pseudo esters of2-benzoylbenzoic acid types. III. J. Am. Chern. Soc., 73, 4627-9. 42. Boehm, H.-P., Diehl, E., Heck, W., and Sappok, R. (1964). Surface oxides of carbon. Angew. Chern.) Int. Ed. Engl., 3, 669-78. 43. Bandosz, T.J.,Jagiello,J., Contescu, C., and Schwarz,J.A. (1993). Characterization of the surfaces of activated carbons in terms of their acidity constant distributions. Ca~on, 31,1193-202. 44. Puri, B.R. and Bansal, R.C. (1964). Studies in surface chemistry of carbon blacks. Part II. Surface acidity in relation to chemisorbed oxygen. C'arbon, 1, 457-64. 45. Puri, B.R. (1970). Surface complexes on carbons. In Chemistry and Physics of Carbon, Vol. 6 (P.L. Walker, Jr, ed.). Marcel Dekker, pp. 191-282. 46. Boehm, H.-P., Mair, G., and Stohr, T. (1985). Oxidative degradation of carbons in alkaline media. Extended Abstracts) 17th Biennial Conference on Carbon) Lexington) KY. The American Carbon Society, pp. 381-2. 47. Contescu, A., Contescu, C., Putyera, K., and Schwarz, J.A. (1997). Surface acidity of carbons characterized by their continuous pK distribution and Boehm titration. Carbon, 35, 83-94. 48. Bashkova, S., Bagreev, A., and Bandosz, T.J. (2002). Effect of surface characteristics on adsorption of methyl mercaptan on activated carbons. Ind. Eng. Chern.) Res., 41,4346-5. 49. Rivin, D. (1962). Hydride transfer reactions of carbon black. In Proceedings 5th Biennial Conference on Carbon) Pennsylvania State University) University Park) PA, Vol. 2. Pergamon Press, pp. 199-209. 50. Zawadski, J. (1989). Infrared spectroscopy in surface chemistry of carbons. In Chemistry and Physics of Carbon, Vol. 21 (P.A. Thrower, ed.). Marcel Dekker, pp. 147-380. 51. Meldrum, B.J. and Rochester, C.H. (1990). In situ infrared study of the surface oxidation of activated carbons in oxygen and carbon dioxide. J. Chern. SOc.) Faraday Trans., 86, 861-5. 52. Ismail, I.M.K. and Walker, P.L., Jr (1989). Detection of low temperature carbon gasification using DSC and TGA. Carbon, 27, 549-59. 53. Zhang, Z.L., Kyotani, T., and Tomita, A. (1989). Dynamic behavior of sutIace oxygen complexes during oxygen chemisorption and subsequent temperatureprogrammed desorption of calcium-loaded coal chars. Energy Fuels, 3, 566-71.
326
Chapter 13 Surface Chemical Characterization of Carbons from Adsorption Studies
54. Burstein, R. and Frumkin, A. (1929). On the behaviour of outgassed activated carbon towards electrolytes (in German). Z. Physik. Chemie (Leipzig) A, 141,21920. 55. Boehm, H.-P. and Voll, M. (1970). Basic surface oxides on carbon - I. Adsorption of acids (in German). Carbon, 8, 227-40. 56. Leon y Leon, C.A., Solar, ].M., Calemma, V., and Radovic, L.R. (1992). Evidence for the protonation of basal plane sites on carbon. Carbon, 30, 797-811. 57. Garten, V.A. and Weiss, D.E. (1957). A new interpretation of the acidic and basic structures in carbon. II. The chromenel carbonium ion couple in carbon. Austral. J. Chem., 10, 309-28. 58. Papirer, E., Li, S., and Donnet, ].-B. (1987). Contribution to the study of basic groups on carbon. Carbon, 25, 243-7. 59. Voll, M. and Boehm, H.-P. (1971). Basic surface oxides on carbon - IV. Chemical reactions for the identification of surface groups (in German). Carbon, 9, 481-8. 60. Garten, V.A. and Weiss, D.E. (1957). Ion and electron exchange properties of activated carbon in relation to its behaviour as a catalyst and adsorbent. Rev. Pure Appl. Chem., 7, 69-122. 61. Contescu, A., Vass, M., Contescu, C., et al. (1998). Acid buffering capacity of basic carbons revealed by their continuous pK distribution. Carbon, 36, 247-58. 62. Suarez, D., Menendez, ].A., Fuente, E., and Montes-Moran, M.A. (1999). Contribution of pyrone-type structures to carbon basicity: an ab initio study. Langmuir, 15, 3897-904. 63. Suarez, D., Menendez,].A., Fuente, E., and Montes-Moran, M.A. (2000). Pyronelike structures as novel oxygen-based organic superbases. Angew. Chem., Int. Ed. Engl., 39, 1320-23. 64. Fuente, E., Menendez, ].A., Suarez, D., and Montes-Moran, M.A. (2003). Basic surface oxides on carbon materials: a global view. Langmuir, 19, 3505-11. 65. Burshtein, R. and Frumkin, A. (1941). Hydrogen peroxide formation in the adsorption of acids by activated charcoal (in Russian). Dokl. Akad. Nauk SSSR, Seriya A, 32,327-9. 66. Matskevich, E.S., Strazhesko, D.N., and Goba, V.E. (1974). Oxidation-reduction properties of carbon in electrolytic solutions. Adsorbtsiya Adsorbenty, 2, 36-9. 67. Darmstadt, H. and Roy, C. (2003). Surface spectroscopic study of basic sites on carbon blacks. Carbon, 41, 2662-5. 68. Puri, B.R., Singh, D.S., Nath, ]., and Sharma, L.R. (1958). Chemisorption of oxygen on activated charcoal and sorption of acids and bases. Ind. Eng. Chem., 50, 1071-4. 69. Lopez-Ramon, M.V., Stoeckli, F., Moreno-Castilla, C., and Carrasco-Marin, F. (1999). On the charactreization of acidic and basic surface sites on carbons by various techniques. Carbon, 37, 1215-21. 70. Biniak, S., Pakula, M., Szymanski, G., and Swiatkowski, A. (1999). Effect of activated carbon surface oxygen groups on adsorption of copper(II) ions from aqueous solutions. Langmuir, 15, 6117-22. 71. Mangun, C.L., Benak, K.R., Economy,]., and Foster, K.L. (2001). Surface chemistry, pore sizes and adsorption properties of activated carbon fibers and precursors treated with ammonia. Carbon, 39, 1809-20. 72. Boehm, H.-P., Mair, G., Stohr, T., et al. (1984). Carbon as a catalyst in oxidation reactions and hydrogen halide elimination reactions. Fuel, 63, 1061-3.
References
32 7
73. Stohr, B., Boehm, H.-P., and Schlogl, R. (1991). Enhgancement of the catalytic activity of activated carbons in oxidation reactions by thermal treatment with ammonia of hydrogen cyanide and observation of a superoxide species as a possible intermediate. Carbon, 29, 707-20. 74. Mang, D., Boehm, H.-P., Stanczyk, K., and Marsh, H. (1992). Inhibiting effect of incorporated nitrogen on the oxidation of microcrystalline carbons. Carbon, 30, 391-8. 75. Noh, J.S. and Schwarz, J.A. (1990). Estimation of surface ionization constants for amphoteric solids. J. Colloid Interface Sci., 139, 139-48. 76. Noh, J.S. and Schwarz, J.A. (1990). Effect of HN0 3 treatment on the surface acidity of activated carbons. Carbon, 28, 675-82. 77. Puri, B.R. (1966). Chemisorbed oxygern evolved as carbon dioxide and its influence on surface reactivity of carbons. Carbon, 4, 391-400. 78. Weiss, A. (1959). Cation exchange properties of clay minerals III. Cation exchange in kaolinite (in German). Z. AnoIg. Allg. Chern., 299, 92-120. 79. Boehm, H.-P. and Schneider, M. (1959). The hydroxyl groups on the surface of the amorphous silicon dioxide "Aerosil" and their reactions (in German). Z. Anorg. Allg. Chern., 301, 326-35. 80. Herrmann, M. and Boehm, H.-P. (1969). On the chemistry ofthe titanium dioxide surface - II. Acidic hydroxyl groups on the surface (in German). Z. Anorg. Allg. Chern., 368, 73-86. 81. Baers, R.F. and Mesmer, R.E. (1976). The Hydrolysis of Cations. J. Wiley & Sons, p. 112. 82. Bansal, R. C., Donnet, J.-B., and Stoeckli, F. (1988). Active Carbons. Marcel Dekker, Chapter 6. 83. Jia, Y.F., Xiao, B. and Thomas, K.M. (2002). Adsorption of metal ions on nitrogen surface functional groups in activated carbons. Langrnuir, 18, 470-8. 84. McDougall, G.J., Hancock, R.D., Nicol, M.J., et al. (1980). The mechanism of adsorption of gold cyanide on activated carbon. J. South African Inst. Mining Metall., 80,344-56.
ADSORPTION ON FULLERENES Fabian Suarez-Garcfa, Amelia Martfnez-Alonso, and Juan M.D. Tasc6n Instituto Nacional del Carbon, CSIC, Oviedo, Spain
Contents 14.1 Introduction 14.2 Adsorption for Porosity Characterization 14.3 Adsorption in the Study of Surface Energetics: Nonreactive Permanent Gases 14.4 Adsorption of Organic Gases and Vapors 14.5 Oxygen Adsorption 14.6 Adsorption Studies using IR Spectroscopy 14.7 Hydrogen Adsorption: Gas Storage 14.8 Adsorption from Solution: Environmental Applications 14.9 Adsorption from Solution: Analytical Applications 14.10 Adsorption from Solution: Colloidal and Biological Systems 14.11 Conclusions Acknowledgments References
32 9 330 33 2 338 34 1 343 346 35 1 353 357 359 359 359
14.1 INTRODUCTION
The first report of the existence offullerenes in 1985 [1], and the subsquent discovery in 1990 of a method to produce them in macroscopic amounts [2], paved the way to a new era of carbon science that involves curved surfaces on the nanoscopic scale. As is well known, the aggregation of fullerene molecules at moderate temperatures and pressures leads to molecular solids termed fullerites. The C 60 (buckminsterfullerene) and C 70 fullerenes and the corresponding fullerites are the easiest to produce, and for this reason they have been the subject of most experimental works. Certain aspects of the solid-state science of fullerenes (e.g., crystal structures, phase transitions, formation of exo- and Adsorption by Carbons ISBN: 978-0-08-044464-2
© 2008 Elsevier Ltd. All rights reserved.
329
330
Chapter 14 Adsorption on Fullerenes
endohedral compounds) relevant to surface studies have been nicely summarized by Bandosz et al. [3] More detailed information can be found in the (already classical) book by Dresselhaus et al. [4]. Adsorption studies on fullerenes have been carried out with a variety of objectives. Thus, in addition to characterizing porosity by means of physisorption, extensive work with either inert gases, organic vapors or even reactive gases as adsorbates has been focused on the characterization of the surface energetics of this type of carbonaceous material. In the case of oxygen, the objective has changed from an initial interest in explaining the high oxidative reactivity of fullerenes to the more recent concern with the effect of oxygen on properties such as electrical conductivity. Although the amount of work devoted to fullerenes has been scarce compared with the interest shown in carbon nanotubes, fullerenes have been studied as hydrogen adsorbents in connection with the storage of hydrogen as a source of energy. In the field of adsorption from aqueous solutions, the applications of fullerenes as analytical tools clearly prevail over other topics such as environmental or biological applications. This disparity in objectives makes the corresponding groups of papers very different from each other. Therefore, in this chapter, we have classified adsorption works on fullerenes according to mixed criteria based on the nature of the adsorbate, the objectives pursued and the methodology followed. This explains the disparity that may exist between the sections that constitute the chapter. The complementary research field where fullerenes constitute the adsorbate rather than the adsorbent has produced interesting results for C 60 adsorption on such materials as zeolite Y [5], activated carbons [6] carbon nanohorns [7], or even clusters of C 60 itself [8] However, this topic is not reviewed here since this book is concerned with solid carbons used exclusively as adsorbents. For similar reasons, we have excluded from this study fullerene-like noncarbon structures such as WS 2 , MoS 2 , NbS 2 , TiS 2 , or BN despite their interest as adsorbents [9] (a review on hydrogen storage on these inorganic nanotubes has been published by Chen and Wu [10]). However, adsorption works on fullerene soot (also termed fullerene black, a solid formed by the condensation of carbon species from the gas phase from which fullerenes are usually extracted) are discussed in view of their possible relevance to fullerenes. Also included in this chapter are simulations of adsorption on schwarzites, due to the close connection between this (still hypothetical) structure and that of the fullerenes.
14.2 ADSORPTION FOR POROSITY CHARACTERIZATION
Soon after the discovery of methods for the mass production of fullerenes, the pyrolysis and gasification behavior of these novel carbons attracted considerable interest due to unexpected findings regarding traditional carbons [11, 12]. In this context, Ismail and Rodgers [11] reported some of the first results on adsorption on fullerene solids. Batches of C 60 of different origin were studied.
14.2
Adsorption for Porosity Characterization
331
Kr, N z, and Oz isotherms at 77 K yielded a low surface area, but CO z adsorption at 298 K indicated that the C 60 crystals studied contained micropores. The solids in each batch studied exhibited different characteristics depending on sample preparation, purification, and age. Kaneko et al. [13] detected microporosity when adsorbing N z on C 60 powder, and interpreted this as being due to the presence of point defects, which had probably been generated by desolvation (i.e., the removal of excess solvent used to extract/purify the fullerene) during heating. Later on, the same team investigated the control of the concentration of defects in C 60 crystals by recrystallization and annealing [14]. N z adsorption isotherms showed the presence of both mesopores (average width of 5 nm) and micropores (average width of 0.8 nm), of which the former disappeared by annealing to rv393 K, whereas the micropores remained even when heating up to 673 K. The authors associated the mesopores with the aggregation of point defects and stacking faults, whereas the micropores were attributed to molecular defects and octahedral vacancy sites. Rathousky et al. [15] found evidence for low-pressure hysteresis in cyclopentane adsorption on a C 60 /C 70 mixture. Later on, they characterized pure C 60 powder by krypton and cyclopentane adsorption [16, 17]. Krypton, which was selected as adsorbent due to the very low surface area of the material, gave rise to a sigmoidal isotherm [16]. Cyclopentane was found to penetrate into the bulk of the C 60 crystals, its presence being detected in the octahedral interstices between the fullerene molecules [17]. Schlagl and coworkers used N 2 adsorption to study the porous texture of several types of fullerene black [18]. The surface areas varied over a rather wide range (11-557 m 2 / g). The fullerene blacks studied contained small amounts of soluble fullerenes. Consequently, the relevance of these results to fullerenes is not much significant. More recently, Beck et al. [19, 20] used N z adsorption to characterize the porosity of fullerene blacks modified by the Diels-Alder reaction. They found that the surface area of micropores increased considerably
after this reaction. Here too, the results bear little relevance to fullerenes since the fullerene blacks studied had been preextracted with toluene to remove the smallest traces of soluble fullerenes. Cascarini de Torre and coworkers [21] measured adsorption isotherms ofN2 , 02' and Ar at 77 K and CO 2 at 298 K on shungite from Zazhoginskoye (Karelia, Russia), a natural carbonaceous material from which fullerenes can be extracted. This rock is made up of a homogeneous distribution of crystalline silicate particles in a noncrystalline carbon matrix. The adsorption results indicated that the material has a low surface area ( rv 25 m z/ g) and an average pore radius of rv 1.7 nm. The gas-solid potential distribution suggested a rather homogeneous surface, with maxima at similar adsorption potentials for N z and Ar. Nagano et al. [22] measured the CO z uptake in C 60 in a study of the effects of supercritical fluid treatment, the aim of which was to remove solvent molecules from C 60 . Carbon dioxide was found to interact strongly with C 60 , and to have a remarkable effect on the orientational phase transition of C 60 crystals at 250 K. The kinetic features of the process suggested that CO 2 absorbs inside the C 60
332
Chapter 14 Adsorption on Fullerenes
solid lattice rather than adsorbing physically on micropores. Later on, Gusev et al. [23] studied nitrogen and argon adsorption on the fullerene C 60 (99.5% purity) and a mixture of (76% C 6o /22% C 70 ), and found no trace of microporosity in their samples. Likewise, Martinez-Alonso et al. found no case of hysteresis or micropore filling in adsorption isotherms of N 2 [24] or Ar (77 K) [25], and the CO 2 (273 K) isotherm they measured was linear [25]. They attributed this lack of microporosity to the high purity and crystallinity degree of the C 60 used. In the above works it was customary to compare novel findings for fullerenes with the behavior ofwell-known carbonaceous solids such as graphite [11], Saran char [11], various types of carbon black [16, 23, 24], orpolycrystalline diamond [16, 24]. Strong similarities were found between fullerene and "graphitized" carbon black for adsorption at low coverages of either Kr [16] and N 2 / Ar [23]. A close similarity was also found between the heterogeneity degrees calculated from the experimental isotherms for C 60 and polycrystalline diamond [24].
14.3
ADSORPTION IN THE STUDY OF SURFACE
ENERGETICS: NONREACTIVE PERMANENT GASES
Some of the papers discussed previously were aimed at characterizing surface energetics rather than porosity, and included theoretical calculations and/or simulations. Thus, Gusev et al. [23] analyzed their experimental data for N 2 and Ar argon adsorption on C 60 using a virial expansion in the Henry's law region and found that, in the low-pressure limit where the fluid-fluid interactions are negligible, the N 2 interaction with the fullerene surface is macroscopically similar to the interaction with "graphitized" carbon black. However, on most of the monolayer the N 2 and Ar affinity for the fullerenes was weaker than for "graphitized" carbon black. Martinez-Alonso et al. [24, 25] combined grand canonical Monte Carlo (GCMC) simulations and experimental isotherms in their studies of adsorption of simple gases (N 2 and Ar at 77 K, and CO 2 at 273 K) on high-purity C 60 . In the simulations, they employed a perfect crystalline structure (face-centered cubic, fcc) of C 60 molecules with lattice parameters and a density that matched the experimental values. The agreement between the simulated and experimental isotherms for all the gases studied was excellent. The adsorption energy distribution functions were calculated from the experimental isotherms, and the energy map corresponding to the model solid was converted into a distribution function and compared with the experimental one. Both distributions agreed quite well, the experimental distribution reflecting all the features exhibited by the distribution of the model solid. The authors deconvoluted the distribution corresponding to the model solid and found three main peaks that matched those of other authors well (see below), even though very different adsorbates and methods of study were used. Ar and CO 2 adsorb in a solid-like phase in the voids of the fullerene solid. The contribution of the "internal" space to the
14.3 Adsorption in the Study of Surface Energetics: Nonreactive Permanent Gases
333
total area was estimated to be 30%. Values for the cross-sectional areas of the gases employed were also given. In a subsequent paper, Tascon and Bottani [26] carried out GCMC simulations for nitrogen adsorption on a defective fullerene, which was created by generating a vacancy in the fcc structure of the perfect solid. The main differences, inferred from simulated nitrogen adsorption, could be ascribed to the difference in surface areas and degree of heterogeneity between the two solids. More recently, the same authors [27] studied ethylene adsorption on C 60 using GCMC simulations. The results validated the simulation model employed and confirmed the assignment of adsorption sites previously reported for other gases. A map of the simulation cell obtained with methane as probe molecule (very similar to that obtained previously with nitrogen) [24] is shown in Fig. 14.1. The C 60 solid exhibits three preferential sites for adsorption: one is located between four fullerene molecules, the second is located in the channels formed between two fullerene molecules, and the third is on the top of the C 60 molecules. The analysis of the adsorption energy distributions with the aid of the gas-gas interaction potential suggested that ethylene is adsorbed in a liquid-like state into the voids of the solid, and that the adsorbed molecules prefer aT-shaped stacking, in agreement with the calculations and experiments reported by other authors [28]. N ext we will discuss a series of theoretical studies from Sandler and coworkers on fullerenes and schwarzites. Following a chronological sequence, we will start with adsorption on schwarzite, a hypothetical structure related to that of the
Figure 14.1 Topographic map ofthe C 60 simulation cell obtained with an ethylene molecule. The X and Yaxes are in arbitrary units and the Z-axis is in angstroms. (Reprinted from Ref. [27] with permission from Elsevier.)
Chapter 14 Adsorption on Fullerenes
334
fullerenes but encompassing convex curvature (either alone, or combined with concave curvature). The interest of these authors in schwarzite was motivated by the possibility of using this and other nanoporous carbons to separate gases of similar dimensions (carbon molecular sieves). Jiang and Sandler chose for their calculations the so-called buckygym C 168 schwarzite [29], and assumed it to be rigid. This model structure has convex and concave surfaces as a result of combining Sp2 and Sp3 hybridizations of carbon atoms, and contains two types of pores with average diameters of 0.7 and 0.9 nm. The pores in the same layer are isolated from each other, but they are connected with those in the neighboring layers by channel intersections. Jiang and Sandler simulated the adsorption of O 2 and N 2 by the GCMC method [30], whereas to simulate the adsorption of an equimolecular 02-N2 mixture they used both this and the Gibbs ensemble Monte Carlo (GEMC) method [31]. Regarding the pure gases, the adsorption isotherms, Henry's constants and isosteric heats of adsorption were calculated for each gas at different temperatures. The calculations showed the dependence of the isosteric heat of adsorption on temperature to be small. For the equimolecular 02-N2 mixture, the GCMC and GEMC methods yielded consistent results. Predictions for the mixture adsorption using the ideal adsorption solution theory (lAST) based solely on the adsorption of pure gases agreed well with the simulation results. The authors of the works just mentioned [30, 31] warned that the accuracy of the Lennard-Jones interaction potentials is critical for accurately computing the properties from molecular simulation, and that inaccurate potentials would lead to large errors. More specifically, they pointed out that the use of parameters such as the Steele potential, which is based on graphite, could lead to errors when calculating the Lennard-Jones interaction potentials, since the effect of surface curvature would not be taken into account. Accordingly, Jiang et al. [32] carried out further calculations for 2, N 2' and a mixture of these two gases in the proportion found in air using two types of potentials for the additive atomatom interaction of each gas with carbon schwarzite. In addition to the Steele potential, they used an ab initio potential obtained from first-principles quantum chemical computations. Their results showed that only the ab initio potential could predict the large adsorption separation the authors expected for 02 and N 2 on C 168 schwarzite. With both potentials, the adsorbed molecules were found to preferentially align along the channel intersection of the schwarzite structure. The predictions for mixture adsorption using the lAST again agreed with the simulations. More recently, Jiang and Sandler carried out similar studies for the adsorption of CO 2, N 2, and their mixture [33] on the same C 168 schwarzite model adsorbent. As an illustration of their results, Fig. 14.2 shows the calculated (competitive) adsorption isotherms, as well as the selectivities of CO 2 over N 2 as a function of pressure for a CO 2-N2 (0.21:0.79) mixture (the composition of this mixture corresponds to the flue gas emitted from the complete combustion of carbon with air). As the isotherms show, the use of the ab initio potential results in a larger difference between the amounts of adsorbed CO 2 and N 2
°
335
14.3 Adsorption in the Study of Surface Energetics: Nonreactive Permanent Gases
8
:§ (5
E
1000
6
C\l
z
B
S
(ti Z v
()
4
N2 C02
en
ab initio Q • Steele 0 •
2
100 C168 (Steele)
oo.-iiiiiGII=tllIE:I:::Q::r:::c::r.dIt:::z:::lit.'c~==~=a;d)
1
10
100
P(kPa)
1000
10000
10
.4-' -0- -
....
-O--Siiicalite
Io..o-I...........~_ _............~-""-..........
1
10
100
....-..-'--l.................".,
1000
10000
P(kPa)
Figure 14.2 Left, adsorption isotherms of the CO 2 -N 2 mixture (bulk composloon CO 2 /N 2 = 0.21:0.79) in the C 168 schwarzite as a function of the total bulk pressure. Right, selectivity of CO 2 over N 2 as a function of the total bulk pressure (bulk composition CO 2 /N 2 = 0.21:0.79) in the C 168 schwarzite (with the Steele and ab initio potentials), silicalite, Na-ZSM-5 (Si/AI = 23), and Na-ZSM-5 (Si/Al = 11). (Reprinted with permission from Ref. [33]. Copyright 2005 American Chemical Society.)
compared with the results obtained using the Steele potential. The selectivities are also higher when calculated with the ab initio potential (values between 100 and 300) than with the Steele one (overall value r-v20). As zeolite membranes have an excellent selectivity to separate CO 2 and N 2 by competitive adsorption [34], the authors also simulated the CO 2 /N 2 competitive adsorption on three types of zeolites, and included the results in Fig. 14.2. These simulations suggest that the SC02 /N2 selectivity for the C 168 schwarzite, predicted by means of either the ab initio or the Steele potential, is greater than that of silicalite, but lower than those of the Na-ZSM-5 zeolites (the increase in selectivity with decreasing Si/AI ratio in the latter zeolite is attributed to an increase in nonframework Na+). The authors therefore concluded that nanoporous carbon adsorbents such as the C 168 schwarzite may be useful for the separation of flue gases. Following their work with schwarzites, Sandler and coworkers carried out theoretical studies on N 2 adsorption at 77 K on C 60 [35] and C 70 [36] using the ab initio-based potential. In the case of C 60 ' Jiang et al. [35] considered the adsorptions on the surface of and within a C 60 crystal separately, these locations yielding type II and type I isotherms, respectively. On the C 60 surface, with increasing pressure, the N 2 molecules were found to sequentially occupy three favorable sites: octahedral ones (between four C 60 ), tetrahedral ones (between two C 60 ) and the top of a C 60 molecule. Therefore, the nature of the sites and the energy sequence agree with results from other teams [24, 27] (see also Sections 14.4 and 14.6). Finally, multiple layers are formed and wetting occurs as the bulk N 2 saturation pressure is reached. Within the C 60 crystal, however, the use of the ab initio potential led to a significantly greater adsorption than the well-known Steele potential. N 2 molecules were observed to intercalate only the octahedral sites, the isosteric heat of adsorption being almost constant.
Chapter 14 Adsorption on Fullerenes
336
Ideal hcp: cIa =1.63
Rhombohedral
fcc
y
Deformed hcp: cIa =1.84
z
~x
Monoclinic
Figure 14.3 Orientational ordering of C 70 in various crystal phases. The semitransparent plane represents the plane along which N 2 was adsorbed in the simulations. (Reprinted with permission from Ref. [36]. Copyright 2005 American Chemical Society.)
In a subsequent study on N 2 adsorption on C 70 ' Arora et al. [36] carried out quantum mechanical calculations to predict N 2 adsorption on five different known structures for C 70 . In this case it has been found that, besides the surface curvature of the C 70 molecule, an additional difference with graphite may arise from changes in the electronic configuration due to the presence of five-membered rings. The surface area, monolayer capacity, and isosteric heat of adsorption were calculated for various C 70 crystal phases [37] that are shown in Fig. 14.3: fcc, deformed hexagonal-closed-packed (hcp I), ideal hexagonalclosed-packed (hcp II), monoclinic (mono), and rhombohedral (rh). Figure 14.4 shows the isosteric heats of adsorption (q~) for N 2 in all of these structures, calculated using both the fluctuation theory and numerical differentiation over a range of loading. The results from both methods are consistent for all crystal types over the entire range of loading. There is a considerable difference in the q~ at the infinite dilution limit as shown in Fig. 14.4. The fcc crystal has the highest value of q~ as a result of its stronger affinity for nitrogen. It can also be seen that the shear-induced phase transformation has a greater effect on the heat of adsorption (substantial drop in q~ between the rh and the hcp II phases) than the orientational ordering transformation (equivalent q~ values for the hcp and mono phases). The isosteric heats of adsorption obtained also indicated that the C 70 fcc crystal surface has a similar affinity for nitrogen to the C 60 fcc crystal surface, both being considerably higher than that of the planar graphite sheets.
14.3 Adsorption in the Study of Surface Energetics: Nonreactive Permanent Gases
337
24 22 20 (5
E -... J
C (;)
0-
•
• fcc iii. hcp I "I' hcp II • mono
•
• rh
18 16 14 12 10 8 0.01
0.1
1
10
(Jlmol/m 2)
Figure 14.4 Isosteric heat of adsorption of N 2 on various C 70 crystal structures at 77 K as a function of loading. The points marked with symbols were obtained from the fluctuation theory, and the solid curves were obtained from numerical differentiation of the configurational internal energy. Reference codes for crystal structures: see text. (Reprinted with permission from Ref. [36]. Copyright 2005 American Chemical Society.)
The adsorption of noble gases on fullerenes was studied by Breton et al. [38] with the objective of defining the conditions that lead to the endohedral or exohedral adsorption of an atom, and also to determine the criteria for the confinement of an atom in a cage. A simple model of interaction potential based on a continuum description of the C 60 molecule was used to describe the encaging properties of alkali metal ions or rare gas atoms. Later on, these authors [39] studied the equilibrium structure of the noble gas-C 6o exohedral complexes (dimer and trimer) on the basis of empirical potentials. They found that the complexes display radial bonds that have magnitudes similar to those between noble gases and aromatic species, and that these complexes behave like floppy supermolecules due to the small corrugation of the fullerene surface. Comparisons were made with equivalent complexes of graphite or benzene instead of C 60 . More recently, Gburski and coworkers used the molecular dynamics method to study systems consisting of a C 60 molecule surrounded by a monolayer or multilayer Ar film [40], as well as exohedral complexes of Ar and Ne that form an ultrathin monolayer film physisorbed on a fullerene surface [41]. Interestingly, since the Ar-fullerene attraction is much stronger than the Ar-Ar attraction, the Ar atoms form a kind of "atmosphere" that surrounds the surface of the C 60 "ball" in the C 60 Ar n conglomerate, which is composed of a fullerene molecule coated with a few dozen argon atoms. Figure 14.5 shows a snapshot of the dynamic, spherically shaped monoatomic layer formed by argons spreading out more or less evenly over the C 60 surface. Szybisz and Urrutia [42] used a physisorption potential to describe the adsorption of 4He inside and outside a single C 60 molecule. They concluded that
Chapter 14 Adsorption on Fullerenes
Figure 14.5 Snapshot of the instantaneous configuration of the Ar atmosphere surrounding C 60 at 48K. Reprinted with permission from Re£ [40]. © 2003, The American Physical Society.
only one 4He atom may be introduced in this fullerene, and that its binding energy is very strong. These authors studied the energetics of the adsorption and determined the structure of the films within the framework of nonlocal density functionals. The evolution toward bulk liquid and surface thickness at the free interface was discussed. In a step forward in the investigation of adsorption of helium isotopes on curved substrates, the same team [43] showed that 3He impurities in sufficiently large 4He systems adsorbed onto substrates with curved geometries form surface bound states. They also found that a single 3He impurity diluted into adsorbed structures such as 4He on the external surface of C 60 behaves as on films on planar substrates and as on 4He pore clusters.
14.4
ADSORPTION OF ORGANIC GASES AND VAPORS
Most work on the interaction of organic vapors with fullerenes has been carried out using inverse gas chromatography (IGC) [44-47]. This is an extension of traditional gas chromatography in which the material to be investigated is packed into a gas chromatographic column, and volatile probe molecules of interest are injected into it. The application of IGC to the study of the surface energetics of different types of carbons has been reviewed by Bottani and Tascon [48]. Abraham et al. [44] determined the gas-solid partition coefficients
14.4 Adsorption of Organic Gases and Vapors
339
for 22 gases and vapors on a C 60 /C 70 mixture using IGC at near zero surface coverage. The probe molecules studied included aliphatic (alkanes, alkyl halides, ethers, esters, ketones, alcohols) and aromatic compounds with different substituents. The results were analyzed using a solvation equation that linearly relates a given property (in this case, the partition constant) ofa series ofsolutes in a fixed phase with different solvation parameters that characterize the solubility properties of the probe vapors (dipolarity-polarizability, 'IT interaction, hydrogen bond formation ability, and dispersion interactions). It was concluded that fullerenes are weakly polarizable and have some hydrogen bond basicity, in agreement with their behavior as "giant closed-cage alkenes" rather than as polyaromatic molecules. The same methodology was used later to study the adsorption of a larger series of organic compounds on fullerene, fullerene-coated surface acoustic wave sensors, graphite, and low-polarity polymers [45]. It was shown that the linear solvation equation calculated for the fullerenes by IGC can be used to determine the relative vapor sensitivities of the fullerene-coated surface acoustic wave sensors. In addition, this equation is useful for comparing the sensitivity of different materials. In all the cases studied in this work, sorption was caused primarily by dispersive interactions. The assembled fullerene layer behaved in a similar way to nonpolar sorbents (graphite and low-polarity polymers) in terms of adsorption selectivity, but yielded less sensitive vapor sensors than linear organic polymers. Davydov et al. [46] used IGC to determine several adsorption thermodynamic properties (equilibrium constants and adsorption heats) for the adsorption of organic compounds on C 60 crystals, and compared them with those obtained for "graphitized" carbon black. The adsorption potential of the surface of fullerene crystals was much lower than that of a carbon black surface. The dispersive interaction of organic molecules with C 60 is much weaker than with carbon black. The adsorption equilibrium constant for alkanes and aromatic compounds is therefore lower in the case of fullerenes. Aliphatic and aromatic alcohols as well as electron-donor compounds such as ketones, nitriles and amines were adsorbed more efficiently on the surface of fullerene crystals. This was taken as proof that fullerene molecules have electron-donor and electron-acceptor properties, which is in agreement with the results of Abraham et al. [44] Papirer et al. [47] showed also by means of IGC that the dispersive component of the surface energy is lower for C 60 than for graphite or carbon black. As an illustration, Fig. 14.6 compares the adsorption energy distribution curves for n-heptane on C 60 , two synthetic carbon blacks (A and B, differing only in the batches), a natural graphite (A) and a synthetic graphite (B). The carbon blacks and graphites show some similarity, with maxima at 18-19 k]lmol (assigned to graphene layers) and 33-34 k]/mol (assigned to adsorption sites located on prismatic planes). For the fullerene, the maximum on the lower-energy side (c. 20 k]lmol) was assigned to graphene-like structures and the second one to oxygenated surface sites; indeed, the occurrence of oxidation was detected by X-ray photoelectron spectroscopy (XPS). No assignment was made to the third peak (29 k]lmol). Therefore, the interpretation of Papirer et al. is not the same as that proposed by other authors [24, 27, 35] (see also
Chapter 14 Adsorption on Fullerenes
34°
0.25
0.20
0.15 0.10
0.05
o 10
16
22
28
34
40
Energy (kJ/mol)
Figure 14.6 Comparison of n-heptane adsorption energy distribution curves determined on C 60 and other carbon materials. (Reprinted from Ref. [47] with permission from Elsevier.)
Section 14.6) for similar three-peak distributions. Specific interactions with cyanomethane, pyridine, chloroform, nitromethane, and i-butanol (polar compounds) exhibited high values for the fullerene, which qualitatively reveals the electron-donor character of C 60 , in agreement with the results of Abraham and coworkers [44]. Chao and Shih [49] used a piezoelectric crystal detection system to study the adsorption of various organic molecules on C 60 . The selectivity of C 60 for polar organic compounds followed the sequence: carboxylic acids>aldehydes> amines> alcohols> ketones. The reversibility of the piezoelectric crystal oscillation frequency during desorption allowed the authors to establish the type of interaction that is produced between organic molecules and fullerene. The behavior of polar molecules must be classified as physical adsorption, except for amines and dithiols, which were chemisorbed. Regarding nonpolar organic molecules, alkynes exhibited a much stronger adsorption on the fullerene than alkenes or alkanes. Chemisorption was also observed to occur in the case of alkynes and 1,3-dienes. The authors concluded that the fullerene-coated piezoelectric quartz crystal can be effectively applied as a detector for various different organic compounds.
14.5 Oxygen Adsorption
34 1
More recently, Hayashi et al. [50] found that toluene can be retained by adsorption on C 6o Pd n , a polymer-like material, at room temperature and at low toluene concentrations. Toluene seems to absorb through its 1T-electrons on partially positive Pd atoms of C 60 Pd n • Theoretical studies have suggested that 1T-electrons of C 60 and toluene overlap through the d-electron orbitals of a Pd atom (thus, not only physical adsorption takes place). This may open a route to fullerene-based materials as adsorbents for harmful volatile organic compounds (VOCs). Other recent studies on organic vapor adsorption on C 60 are connected with either applications in chromatography [51], or as a reference for comparison with carbon nanotubes [52]. Mixteco-Sanchez and Guirado-Lopez [53] carried out semiempirical (MNDO) and ab initio (density functional theory) calculations for the structural and electronic properties ofthiol [SCH3 and S(CH3)2CH3] molecules adsorbed on C 60 and various types of carbon nanotubes. The results showed that, in the low-coverage regime, the adsorbed thiols prefer to aggregate as a cluster on one side of the C 60 cage, something which, according to the authors, could be of fundamental importance for the synthesis of C 6o -Langmuir monolayers in specific environments. With increasing coverage, increasing repulsion desestabilizes the molecular bundle and a transition to a more unform distribution is achieved. The authors also observed considerable distortions of the spherical carbon structure upon thiol adsorption, which clearly demonstrate the considerable strain to which fullerene materials may be subjected. Turning our attention now to studies on the by-products of producing fullerenes as adsorbents for VOCs, a so-called fullerene-type deposit was investigated as an adsorbent in connection with its possible use in organic compound gas chromatographic separation [54]. At the same time, a fullerene-extracted soot was studied as an adsorbent for collecting VOCs in ambient air [55]. In more recent works, the adsorption of organic vapors such as benzene and ethanol on fullerene blacks [56, 57] has been compared with that of permanent gases (Ar, N 2 , CO 2 ).
14.5
OXYGEN ADSORPTION
In 1992, Rao and coworkers [58] reported pioneering experimental work by XPS on oxygen and nitrogen adsorption on C 60 films. The interactions of both gases with the fullerene were strong, and a sharp feature around 400.6 eV in the case ofN2 at 80 K was attributed to a strongly chemisorbed molecular species. In the case of oxygen, reactive interaction with the formation of oxygenated C 60 was found to occur without any special irradiation treatment. The same authors [59] also found strong interactions of fullerenes with transition metals such as Cr, Ni, and Cu deposited on the C 60 films. Schlagl and coworkers were also among the first to investigate C 60 interaction with oxygen [60], although in connection with degradative chemical reactions
Chapter 14 Adsorption on Fullerenes
34 2
o
;
~ Intercalation
Peroxides?
Insoluble polymer
Opened cages
Reactive intermediates
Figure 14-7 Reaction pathway for the thermally induced oxidation of C 60 with molecular oxygen. (Reprinted from Ref. [61] with kind permission of Springer Science and Business Media.)
(alluded to above [11, 12]) rather thanjust adsorption. Later on, they summarized the results ofoxidation experiments on solid C 60 and related them to intercalation and de-intercalation experiments with 02' CO, and CO 2 [61]. The intercalated species were characterized by temperature-programmed desorption (TPD) and infrared (IR) spectroscopy. Figure 14.7 shows a scheme proposed by these authors for the reaction pathway for C 60 with oxygen. It has been found that the intercalation/de-intercalation process of molecular oxygen in C 60 formed clathrates in the interstitial voids of the C 60 lattice. Therefore, once exposed to the oxygen, C 60 samples cannot be retrieved without changes to their original form. The C 60 material either contains intercalated oxygen or is to some extent oxidized or polymerized due to the heat treatment necessary for removing the oxygen. In the light of these results, the authors acknowledged that C 60 could not be regarded as an appropriate host lattice for the intercalation of oxygen [60]. More recently, Wu et al. [62] studied oxygen adsorption on the surface of Rb 6 C 60 films in connection with the superconducting properties of this and other alkali fullerides, as their electrical conductivity disappears after they are exposed to air. XPS and ultraviolet photoelectron spectroscopy showed that oxygen first adsorbs rapidly to form a peroxide on the top surface. After most of the Rb atoms intercalated in the C 60 film have moved to the surface, a linear oxygen uptake occurs together with the formation of carbonate and superoxide species. Also in connection with electrical conductivity, Tanaka et al. [63] carried out a study on the semiconducting properties of C 6o /zeolite Y and K-C 6o lzeolite Y systems and their dependence on a gas atmosphere. Oxygen
14.6 Adsorption Studies using IR Spectroscopy
343
adsorption experiments under UV irradiation have resulted in both rapid and slow current decays with time, indicating that the photogenerated carriers are mainly electrons, with a small contribution from holes. The slow current decay has been interpreted as the diffusion of oxygen into zeolite pores. The authors also pointed out that ethylene sensing on C 6o /zeolite is possible in the dark and follows the behavior of a Langmuir-type isotherm, which is attributable to the compression of ethylene molecules into zeolite pores. Matsumoto et al. [64] investigated gas occlusion in C 60 crystals by spectroellipsometry. They found that some oxygen remains in the voids (associated with polycrystals) in chemisorbed form. This is in contrast with the behavior of C 60 crystals in the presence of He, Ar, H 2 , or N 2 , where the spectra changed reversibly with pressure, in accordance with a physisorption model whereby gas molecules enter the voids and are occluded as a quasiliquid. Niklowitz et al. [65] studied the interaction of oxygen molecules with a fullerene surface using electron energy loss spectroscopy and TPD. On the basis of the vibrational excitation behavior, the authors concluded that molecular oxygen was physisorbed on C 60 under the conditions studied (20 K). In other words, the adsorbed molecules were only weakly perturbed by the C 60 substrate.
14.6
ADSORPTION STUDIES USING
IR
SPECTROSCOPY
In this section, we will review a series of papers [66-72] on the adsorption of several gases with different reactivities (carbon and nitrogen oxides, light hydrocarbons, alcohols) on C 60 fullerene. The works discussed here were produced by a single team (Folman and coworkers) and have in common the experimental approach used, viz. the study of the species adsorbed on C 60 films by IR spectroscopy. Some other papers dealing with the adsorption of some of these gases (C0 2 , light hydrocarbons) on C 60 have been discussed elsewhere in this chapter as they bear a closer relation (in experimental methodology or objectives) to the topics treated in those sections. As early as 1992, Fastow et al. [66] reported the IR results for CO and NO adsorption on C 60 . The spectra recorded at 77 K for CO adsorbed on C 60 films, which have led to further analyses being carried out, are reproduced in Fig. 14.8. Two partially overlapping absorption bands, positioned at 2135 ± 1 and 2128 ± 1 cm- 1 , can be observed. The appearance of two bands suggests that CO adsorbs on two different sites on the C 60 surface. The large spectral shifts regarding gas-phase frequency (2153 em -1) indicate that the interaction is relatively strong. A similar conclusion was drawn from NO adsorption, which also showed a multiplicity on the two absorption bands, an indication that NO is adsorbed on its dimer form in two different sites. Shortly after this work, IR spectra for CO 2 and N 2 0 adsorption on C 60 , graphite and diamond films were also studied [67]. In this case, only one adsorption site was
344
Chapter 14 Adsorption on Fullerenes
2200
2150 2100 Frequency per (cm)
2050
Figure 14.8 IR spectra of CO adsorbed on C 60 • The different absorption bands correspond to different coverages. (Reprinted with permission from Re£ [66]. Copyright 1992 American Chemical Society.)
detected on the three allotropes. Large spectral shifts and long desorption times were recorded for both the CO 2 and N 2 0 adsorbed on C 60 in comparison with the graphite and diamond studied. All these findings, combined with the available literature for CO and NO adsorption, indicated a strong interaction of CO, CO 2 , N 2 0, and NO with C 60 compared to the two other carbon allotropes. In another study, Heidberg et al. [68] investigated CO and CO 2 adsorption under ultrahigh vacuum conditions on C 60 thin films deposited on KBr(100). The IR spectra were recorded at different polarizations. The results obtained with both adsorption systems again point to the existence of two different adsorption sites on the C 60 film. The spectra showed different absorption intensities depending on the type of polarization. This was interpreted by the authors as being due to the anisotropy of the C 60 film. From the studies discussed hitherto [66, 68], it has been assumed that the two CO adsorption sites at 77 K on the C 60 surface could be on top ofa C 60 molecule, and in voids between these molecules. To obtain further information on the
14.6 Adsorption Studies using IR Spectroscopy
345
nature of these sites, calculations of adsorption potentials and spectral shifts were made by Fastow et al. [69] and Folman et al. [72] using the Buckingham-Corner (six-exponential) and the Lennard-Jones potentials. A number of adsorption sites were chosen, including the void space between four, three, and two neighboring C 60 molecules on their respective surface planes, and the center of the hexagon and pentagon on the surface of a single C 60 molecule. The potentials calculated clearly indicated that the adsorption sites in the voids between the C 60 molecules are energetically preferred to sites on top of the C 60 molecules (for the latter sites, higher potentials and lower spectral shifts were obtained). In turn, the calculated spectral shifts for the sites between four (-15 cm -1) and two (-8 cm -1) C 60 molecules were very similar to the experimentally measured values, suggesting that indeed those are the preferred sites. The results of the work of other teams [24, 25, 27, 35] later on were found to agree with this interpretation on the nature of the adsorption sites on the C 60 surface, despite the significant differences in probe molecules and measurement techniques used. The nature of the adsorption sites for CO on C 60 gave rise to yet another work, in which Lubezky et al. [71] used LiFjC 60 and NaCljC 60 mixed films as adsorbent in an attempt to obtain spectra of CO adsorbed on individual C 60 molecules which might be present in the films as a result of their simultaneous deposition from two separate sublimation sources. The results obtained with the LiF j C 60 films fulfilled their expectations: apart from two bands ascribable to CO adsorption on LiF, a third band at 2130 cm -1 could be ascribed to CO adsorption on C 60 . This clearly showed that CO adsorbs on top of individual C 60 molecules dispersed in the LiF matrix. With NaCljC 6o , however, no IR bands for CO adsorbed on C 60 were found. Isosteric heat calculations revealed that the isosteric heat of CO on NaCI (18.8 kJ/mol) is considerably higher than that on single C 60 molecules (11.7 kJ/mol). Therefore, preferential adsorption is thought to take place on NaCI, which would explain the absence of bands for CO adsorbed on C 60 . Finally, the same team [70] also studied the IR spectra of light hydrocarbons, methanol and their deuteriated counterparts adsorbed on C 60 films at different temperatures. In the case of methane, deuteriated methane, ethylene, and acetyene, shifts in the frequencies of the adsorbed molecules compared to the gas phase were found. The shifts were larger for C 2 H 4 and C 2 H 2 than for CH 4 , which was attributed to the occurrence of a higher dispersion interaction and possibly 1T-1T interactions for the former gases. Adsorbed methanol gave an O-H stretching vibration band at a frequency (3320cm- 1 ) similar to that of liquid methanol, suggesting that CH 3 0H adsorbs in the form of clusters. In addition, a strong band at 1028 cm- 1 (attributed to the c=o stretching vibration) persisted in part and was shifted to 1024 cm -1 at higher temperatures and on evacuation. This suggested that a small fraction of methanol was either chemisorbed, or physisorbed on high-energy sites. Similar results were obtained when deuteriated methanol was adsorbed on C 60 .
Chapter 14 Adsorption on Fullerenes
346
14.7
HYDROGEN ADSORPTION: GAS STORAGE
Interest in hydrogen as a fuel has increased very sharply in recent years due mainly to advances in technologies for hydrogen production and utilization. However, it is also necessary to develop efficient systems for storing it before the mass-scale use of hydrogen as a fuel can be achieved. The strategy elaborated by the US Department of Energy (DOE) requires that a weight efficiency of 6.5 wt% and a volumetric density of 62 kgH 2 /m 3 at room temperature be reached before hydrogen can be used as a potential source of energy. Numerous studies have been published in recent years on the use of different carbon materials as adsorbents for hydrogen storage. These include activated carbons as well as novel carbon forms such as carbon nanofibers, multi-wall carbon nanotubes (MWNTs), single-wall carbon nanotubes (SWNTs), and also fullerenes. Much work on the reversible adsorption of hydrogen was stimulated by results published by Dillon et al. [73] In this work, TPD was used to measure the amount of hydrogen desorbed from soots containing around 0.1-0.2 wt% open SWNTs (as estimated by transmission electron microscopy). The amount of hydrogen desorbed at a peak at c. 300 K corresponds to a gravimetric storage density per SWNT of 5-10 wt%. According to the authors, the rest of the carbon material (> 99 wt%) is not thought to take part in hydrogen adsorption. Since then, a number of papers [73-77] and reviews [78, 79] have been published on the use of new carbons for hydrogen storage. Among the problems referred to were the wide dispersion of results and the lack of reproducibility. The discrepancies identified may be due to the use of low purity materials (especially in the case of nanotubes) or experimental errors, especially when hydrogen is used at high pressures [76] and, in the case of theoretical studies, unrealistic models. In principle, fullerenes could meet some of the requirements established by the DOE. Thus, C 60 is theoretically able to store 7.7wt% H 2 assuming the bonding of one hydrogen atom per one carbon atom (chemisorption), which would lead to the formation of C 6o H 60 [79]. In a recent survey of possible states for hydrogen in the hydrofullerene, Schur et al. [80] identified at least two states: lattice and fullerenated hydrogen (exo and endo, respectively). Lattice hydrogen is present in the form of solid solution and is distributed in the interstitial sites of the lattice of fullerenes (fullerite). Depending on the type of cubic lattice of the fullerite (fcc, or body-centered cubic (bcc)) the maximum hydrogen content in the lattice per fullerene molecule is two atoms (C 6o H 2 ) for the fcc and six atoms (C 6o H 6 ) for the bcc (see Fig. 14.9). According to the authors, these structures are stable below 293 K. The other possible state for hydrogen, the fullerenated state, contains hydrogen atoms that are chemically bound with the carbon atoms forming the fullerene cage. Figure 14.10 shows a unit cell of hydrofullerite with an fcc structure, where both the lattice and fullerenated hydrogen can be seen. Fullerenes exohydrogenated to different degrees have also been experimentally prepared, the
14.7 Hydrogen Adsorption: Gas Storage
(a)
347 (b)
Figure 14.9 Unit cells of fullerite with fcc (a) and bcc (b) structures with lattice hydrogen only. (Shaded circle) Sites ofcrystalline lattices, in which fullerenes molecules are distributed; (.) octahedral interstitial sites; (0) tetrahedral interstitial sites, in which atoms of lattice hydrogen are located. (Reprinted from Ref. [80] with permission from Elsevier.)
Figure 14.10 Unit cell of hydrofullerite with an fcc structure, containing both lattice and fullerenated hydrogen. (Shaded circle) Sites of crystalline lattice, in which hydrofullerenes are distributed; (.) octahedral interstitial sites; (0) tetrahedral interstitial sites, in which atoms of lattice hydrogen are located. (Reprinted from Re£ [80] with permission from Elsevier.)
most stable being C 6o H 6 , C 6o H 18 , and C 6o H 36 • Figure 14.11 shows a molecule of exohydrogenated C 6o H 60 ' The process of hydride formation in the fullerites consists of two steps: saturation of the fullerite lattice with mobile hydrogen and the hydrogenation of fullerene molecules by excess mobile hydrogen. There are various methods for
348
Chapter 14 Adsorption on Fullerenes
Figure 14.11 Exohydrogenated molecule of fullerene. (.) Carbon atoms forming the molecule cage; (0) atoms of fullerenated exohydrogen. (Reprinted from Ref. [80] with permission from Elsevier.)
the synthesis ofhydrofullerenes, such as reaction with gaseous hydrogen (direct, or metal-catalyzed), Birch reduction, hydrogen transfer reactions, and others. From a practical point of view, gas-solid processes would seem to be the most appropriate options. C 60 hydrogenation through Birch reduction (Li, liquid NH 3 , tert-BuGH) was reported by Haufler et al. [81], who identified (using mass spectrometry, 1 H nuclear magnetic resonance and IR spectroscopy) C 6o H 36 and C 6o H 18 as reaction products, although it was not posible to determine whether the latter was the result of Birch reduction or C 6o H 36 pyrolysis. The authors studied the dehydrogenation by treating a solution of hydrofullerene in toluene with dichloro dicyano quinone under reflux. Thin-layer chromatographic and mass spectrometric analyses showed that the dehydrogenated material was exclusively C 60 , and led the authors to conclude that the reaction of fullerenes could be totally reversible with no alteration to the molecular skeleton occurring during the Birch reduction. Direct, noncatalyzed hydrogenation of fullerenes at high pressures has been reported by different authors. The methodology followed and the degree of hydrogenation achieved vary widely. Kolesnikov et al. [82] studied samples of fullerite hydrogenated at 3 GPa, and found that the resulting material consisted of C 6o H 32 molecules with molecular hydrogen dissolved on interstitial sites. No results on the dehydrogenation were reported. Ye et al. [83] studied the adsorption and desorption of hydrogen at 12 MPa and 77 K on different fullerite samples (with an approximate composition of75% C 60 and 22% C 70 ) as well as
14.7 Hydrogen Adsorption: Gas Storage
349
on pure C 60 and C 70 . Pure (>99.9%) C 60 adsorbed 0.83 wt% Hz and desorbed 0.70 wt%. One of the fullerite samples exhibited a maximal hydrogen adsorption of 4.4wt% and a desorption of up to 4.38wt% Hz. Jin et al. [84] attempted to determine whether treating C 60 with hydrogen at a high pressure (65 MPa) and temperature (573 K) would suffice to provoke the opening of the fullerene cage and the entry of hydrogen molecules. Mass spectroscopic analyses revealed the formation of C60HZ-18, corresponding to hydrogens bound exohedrally to the fullerene cage. Apparently, therefore, no access of hydrogen to the inner part of the cage takes place under these conditions. Meletov et al. [85] carried out fullerene hydrogenation at 3 GPa, between 650 and 700 K for different durations of time. The main product obtained (95%) was C 6o H 36 , the remaining 5% being fullerenes hydrogenated to smaller extents. No data on desorption were provided since this work was aimed at studying the formation of different isomers of C 6o H 36 by Raman spectroscopy. Kurmaev et al. [86] also carried out the hydrogenation at 3 GPa, between 620 and 770 K, and for different durations of time. The product obtained was C6oH37.S-46.S (as determined by elemental analysis). Again, no desorption results were presented as the main objective of this work was to study hydrofullerene isomers by X-ray fluorescence. Tarasov et al. [87] tried to hydrogenate (deuterate) fullerites (85% C 60 + 15% C 70 + 2% of higher fullerenes) at moderate pressures (1-2.5 MPa) by mixing the fullerite with intermetallic compounds (LaNi s ' LaNi4.6sMno.3s, and CeC0 3) or metals (V and Pd). Hydrogenation does not occur at room temperature, it being necessary to work in the range of 573-673 K. In addition, several hydrogenation cycles are necessary to obtain the maximum hydrogen (deuterium) content. Thus, C6oDz4-Z6 was obtained after seven cycles at 2.5 MPa and 673 K. These authors studied the decomposition of the hydrogenated materials by means of differential thermal analysis and thermogravimetry. A first peak at 350-600 K was attributed to hydrogen desorption from the metal hydrides. A second peak at 800 K probably corresponds to fullerite dehydrogenation. At higher temperatures, the fullerene structure decomposes giving rise to metal carbides, except in the case of Pd where no peaks are observed above 800 K. In contrast to these results, Brosha et al. [88] showed that dehydrogenation occurs alongside the decomposition of the fullerenes. These authors carried out the hydrogenation of C 60 (direct hydrogenation) and C 6o Ru 3 (catalyzed hydrogenation) at 0.3 MPa and 673 K, giving rise to C6oH18.7 and C6oRu3Hz4' respectively. The dehydrogenation was monitored by means of thermogravimetry, which showed that the samples are stable up to 703 K. Above 727 K, mass loss occurs due to dehydrogenation, accompanied by the destruction of fullerenes (these authors observed the evolution of methane besides that of Hz). The X-ray diffraction patterns of the dehydrogenated samples did not correspond to C 60 , but to an amorphous carbon material. Another method for preparing hydrogenated fullerenes is that of transfer hydrogenation [89-93]. This consists of transferring hydrogen from 9,10-dihydroanthracene to C 60 at 623 K, to yield mainly C 6o H 36 , but sometimes
350
Chapter 14 Adsorption on Fullerenes
accompanied by C 6o H 18 • Most works in this regard have focused on the determination and characterization of the isomers of C 6o H 36 that are formed. However, in a work by Dorozhko et al. [91], the temperature evolution of C 6o H 36 and C 6o H 18 was studied. It was concluded that the isothermal treatment of C 6o H 36 at 660 K leads to C 6o H 18 as an intermediate product. As the temperature increases to 700 K, dehydrogenation takes place and an increase in the mass spectrometry peak corresponding to C 60 is observed. Nevertheless, differences were observed between the IR spectra of the hydrogenated samples and the original C 60 , which led the authors to conclude that no pristine C 60 was recovered under the conditions of their experiment. An additional possibility is that of endohydrogenated fullerenes. The formation of these compounds implies that hydrogen must cross the rings in the fullerene cage. Figure 14.12 illustrates this possibility, with hydrogen molecules located inside the cage [94]. Endohedral hydrogen adsorption has been approached theoretically [94-97], and would seem to offer little chance for practical application as the desorption could only be produced by the rupture of the fullerene cage [80]. However, Narita and Oku [95] point out that the energy required for the discharge ofH 2 from fullerene materials is similar to that of H 2 storage. These authors simulated H 2 storage in C 60 by means of molecular
Figure
14.12 Endohedral structures with various amounts of hydrogen molecules (from 9 to 24) inside the C 60 cage. (Reprinted from Ref. [94] with permission from Elsevier.)
14.8 Adsorption from Solution: Environmental Applications
35 1
dynamics calculations, and concluded that the H 2 molecules are in a stable state inside the C 60 cage at 298 K and 0.1 MPa. They also confirmed that a pressure of over 5 MPa is required to store H 2 molecules in a C 60 cage, and that H 2 molecules can enter through the hexagonal rings in the fullerenes. Barajas-Barraza and Guirado-L6pez [96] analyzed the hydrogen storage behavior in spheroidal C 60 and C 82 , as well as in cylindrical finite-length (5, 5) armchair C and BN fullerenes, by means of semiempirical (MNDO) as well as ab initio density functional theory calculations at T = 0 K. They observed that, whereas chemisorption of hydrogen individual atoms can be produced on the external surface of fullerenes, the hydrogen atoms cannot be bound to any internal surfaces. Therefore, hydrogen can only exist in molecular form inside the fullerenes. The maximum amount of hydrogen that can be stored inside a C 60 molecule is 23 molecules. This maximum storage capacity is in good agreement with the result reported by Tiirker and Erkoc; [94] (24 hydrogen molecules). The latter study was carried out by means of the AMl self-consistent field molecular orbital method at the restricted Hartree-Fock level. These authors pointed out that all the systems nH 2 @C 60 (n: 9, 12, 15, 19, 21, 24) studied are stable but highly endothermic. In a recent work, Oksengorn [98] presented an experimental procedure for preparing endohydrogenated fullerenes, whereby a beam of light with A = 532 nm (from an Nd-Y AG laser) is used to excite C 60 in the presence of hydrogen at a pressure of 0.1 GPa. The fraction of C 60 that was hydrogenated contained 18% endohydrogenated fullerene, an amount higher than those produced in previous attempts to produce endohydrogenated fullerene. Before concluding this section, we would like to mention a theoretical work by Turnbull and Boninsegni [99] on p-hydrogen adsorption on the outer surface of a single fullerene. Monte Carlo simulations showed that a single solid monolayer is thermodynamically stable, and is commensurate with the corrugated surface of the fullerene. As the chemical potential was increased, a discontinuous change to an inconmensurate, compressible layer was observed. No evidence for quantum exchanges between the p-H 2 molecules was observed.
14.8
ADSORPTION FROM SOLUTION: ENVIRONMENTAL ApPLICATIONS
Studies of fullerenes as adsorbents from an aqueous solution in relation with the removal of pollutants are relatively scarce. Thus, Berezkin et al. [100, 101] investigated the adsorptive activity of fullerenes for organic pollutants in water and compared them with activated carbon and soots with and without fullerenes. They studied the purification of natural river water and waste from a pharmaceutical plant. The latter liquid contained various aliphatic, cyclic and aromatic compounds, the overwhelming majority of organic impurities being chlorinated compounds. The adsorption behavior ofsoot was found to be similar
Chapter 14 Adsorption on Fullerenes
35 2
to that of activated carbon, but fullerenes were more efficient than these two sorbents. The authors concluded that adsorption on fullerenes proceeds mainly by physical adsorption through dispersive interaction forces (it is worth recalling at this point that Abraham and coworkers [45] proposed that VOC adsorption on fullerenes also takes place through dispersive interactions). Berezkin et al. [101] also found a correlation between the adsorption properties of fullerenes and specific features of their solubility, and attributed the existence of this correlation to the action of the same intermolecular forces when interacting with the same molecules of adsorptives or solvents. Cheng et al. studied the interactions between C 60 and two common environmental contaminants, naphthalene [102, 103] and 1,2-dichlorobanzene [103]. Both adsorption and desorption were studied using C 60 either deposited as a thin film, or dispersed in water. Enhanced dispersion of C 60 in water (which was attained by causing the disaggregation of C 60 particles) was found to increase the extent of organic pollutant adsorption by several orders of magnitude. As Fig. 14.13 shows, a strong adsorption/desorption hysteresis effect could be
12000"
(a) 0 0 10000 OIIIJI/I·O
()o
8000 'C)
0> 6000 ~ 0-
4000 "
•
2000
•
•
0 0.1
0
0.2
0.3
0.4
0.3
0.4
CwOlg/ml) 12000 (b) 10000 8000 'C)
.........
0>
~
6000
0-
4000
•
2000 0 0
0.1
0.2
Cw (flg/ml)
Figure 14.13 Adsorption-desorption isotherms of naphthalene on C 60 (plots a and b correspond to two different samples of "C 60 small aggregates"). Solid diamonds, adsorption data; empty diamonds, desorption data; solid line, Freundlich isotherms (fitted with the adsorption data). (Reprinted with permission from Ref. [102]. Copyright 2004 American Chemical Society.)
14.9 Adsorption from Solution: Analytical Applications
353
observed (the authors ascertained the accuracy of equilibrium). The authors explained this occurrence of hysteresis using a "two-compartment" model, whereby adsorption takes place first on the external surface that is in contact with water (this adsorption being reversible), irreversible adsorption occurring on the internal surface inside the aggregates of C 60 molecules.
14.9
ADSORPTION FROM SOLUTION: ANALYTICAL
ApPLICATIONS
In their review of the impact of fullerenes in analytical sciences, Valcarcel and coworkers [104] identified two main connections between fullerenes and analytical chemistry. The first relationship views fullerenes as analytes. This would involve quantifying them in various types of samples, such as biological tissues (in relation to the possible toxicity of C 60 ) or geological materials (shungite, fulgurite). The second relationship attributes to fullerenes the function of analytical tools, namely as chromatographic stationary phases, as electrochemical and optical sensors, or as sorbent materials in continuous flow systems. These three alternatives are sketched in Fig. 14.14. Interestingly, adsorption interactions playa key role in all of these analytical systems. Here we will briefly outline the relationship between adsorption on fullerenes and these analytical applications, taking advantage of the framework provided by the review ofBaena et al. [104]. Let us first consider adsorption in relation with fullerene applications in liquid chromatography (LC) and high-performance liquid chromatography (HPLC). Saito et al. [105] first used a chemically bonded C 6o -silica as a stationary phase for LC, and found it to have a selectivity different from that of more traditional
~:" . '" ,."' . ~
Chemical sensor
GC LC
Figure 14.14 The three main possibilities for direct use of fullerenes in analytical processes. MP, mobile phase; I, injector; D, detector; GC and LC, gas and liquid chromatographs. (Reprinted from Ref. [104] with permission from Elsevier.)
Chapter 14 Adsorption on Fullerenes
354
(a)
(b)
a .....----- 43.4A - - - - -...
Figure 14.15 Molecular modelling scheme for interaction between PAHs and C 60 bonded silica phases by space filling model. (a) triphenylene and o-terpenyl on the C-high type C 60 phase. (b) o-terpenyl on the C-Iow type C 60 phase. (Adapted from Ref. [106] - reproduced by permission of the Royal Society of Chemistry.)
octadecylsilyl silica (ODS) phases. A good correlation existed between the retention data for polyacyclic aromatic hydrocarbons (PAHs) with this C 60 bonded phase, and with C 60 itself as a stationary phase. In another work, the same team [106] synthesized various chemically bonded C 6o-silica phases, and investigated systematically the retention behavior of PAHs. They found that the chemically bonded C 6o -silica phases showed a higher selectivity for aromatic compounds and planar molecules that resulted in a unique molecular recognition capability for PAHs. The C 60 bonded phases with high surface coverage (termed C-high type) had a selectivity different from that of phases with low surface coverage (termed C-low type) for the separation ofisomeric PAHs. To explain these findings, the authors proposed a molecular model, which is reproduced in Fig. 14.15. The cavities between the closely bonded ligands of the C-high type C 60 phases are smaller. Nonplanar molecules such as o-terphenyl may be expected to have more difficulty in penetrating these cavities than planar molecules such as triphenylene. This would lead to a decreased 1T-1T interaction of the surface with C 60 for the nonplanar molecules. In this case, the planar molecules would be retained longer than the nonplanar molecules. In contrast, the C-low type C 60 phases contain larger cavities between less densely bonded ligands (Fig. 14.15). This kind of cavity can readily receive both nonplanar and planar molecules. Therefore C-low type C 60 phases cannot be expected to discriminate between PAH isomers in terms of planarity and will produce smaller retentions for PAHs than the C-high type. The same type of selectivity was found by Stalling et al. [107] in the HPLC separation ofpolychlorinated biphenyls (PCBs) using a C 6o / 7o -polystyrene divinylbenzene bonded phase. This material acted as an electron donor-acceptor adsorbent which provided enhanced enrichment of coplanar PCB constituents,
14.9 Adsorption from Solution: Analytical Applications
355
including chlorinated dibenzofurans and dibenzo-p-dioxins. Bianco et al. [108] developed another HPLC stationary phase based on a fullerene derivative covalently linked to silica gel microparticles. Exceptionally high size selectivities were obtained for cyclic oligomeric compounds such as calixarenes and cyclodextrins. In this new phase, solute dimensions dictated the retention behavior of macrocyclic compounds, while shape and functionality modulated the relative retentions of a series of protected peptides. In the field of gas chromatography (GC), Golovnya et al. [109] first developed a fullerene-based stationary phase (consisting of a C 60 coating on a capillary glass column) that was used for the retention ofhigh-boiling organic compounds, such as aromatic and aliphatic hydrocarbons, amines, alcohols, and esters. Later on, fullerene-containing polysiloxanes were developed for use as stationary phases in GC [110-112]. These phases were highly suitable for the separation of high boiling point compounds like PAHs and phthalic esters. PAHs were eluted in the sequence of their increasing dispersion force, and the methyl esters of the unsaturated acids were eluted after the corresponding esters of saturated acids due to '1T-'1T interaction of fullerene with the double bond of the methyl ester of the unsaturated acid. Therefore, as in the case ofLC, the adsorption mechanism is attributed to strong '1T-'1T interactions and donor-acceptor interactions of fullerene with analytes. This close similarity in mechanisms involved in LC and GC prompted us to include here some discussion on the latter technique despite the fact that, in GC, adsorption takes place from the gas phase rather than the liquid phase. Fullerenes as sensors is the second field of analytical applications considered in Baena et al.'s review [104]. Some studies relative to the use of fullerenes as sensors for organic vapors have already been reviewed in Section 14.4 [44, 45, 49]. Following their inclusion of a fullerene-coated piezoelectric (PZ) quartz crystal membrane to study the adsorption of various organic molecules from the gas phase [49], Shih et al. [113] developed several applications of this type of sensor to organic and inorganic species in solutions. Thus, it was found that a C 6o -cryptand22-coated piezoelectric crystal sensor could be used as an LC detector not only for organic molecules but also for metal ions in solutions. In the field of organics, the C 6o -cryptand22-coated LC-PZ system compared well in performance with a commercial UV-VIS detector. Moreover, cryptands are well-known for their remarkable complexing ability for both cations and anions and, accordingly, the C 6o -cryptand22-coated LC-PZ detector has been shown [113] to act as a switch-type multifunctional detector that can be used either as a cationic detector at pH 2: 7 or as an anionic detector at pH ~ 6. Other applications of fullerenes as sensors [114-116] bear little or no connection with adsorption phenomena. The third field of analytical application of fullerenes identified by Baena et al. [104] is as phases for cleaning and preconcentrating analytes. Practically all of the work carried out in this field with fullerenes was carried out by one laboratory. As early as 1994, Gallego et al. [117] first reported on the analytical potential of C 60 fullerenes as sorbent materials for the preconcentration
356
Chapter 14 Adsorption on Fullerenes
of metal traces through the formation of neutral chelates. The model system they used was to determine Pb traces in waters by using ammonium pyrrolidine dithiocarbamate as a ligand. The chelate is formed in a continuous flow system, adsorbed on a C 60 fullerene microcolumn, and subsequently eluted for transfer to an analyzer (e.g., an atomic absorption spectrometer). The authors determined the adsorption isotherms of Pb at low concentrations, and found that C 60 had a greater adsorption capacity than ODS and activated carbon, which were tested as sorbents under equivalent conditions. The value of the Freundlich K constant was maximum for C 60 , indicating that the van der Waals interactions between the ligand (chelate) and the fullerene are stronger than with the other two sorbents. Therefore, of the three sorbent materials studied, C 60 was considered the most effective for the preconcentration of traces oflead thanks to its higher adsorption capacity. More importantly [104], C 60 exhibited the highest selectivity. Subsequently, the same team using C 60 and/or C 70 as sorbents for preconcentratation in continuous systems were able to determine Cd [118], Cu [119], Cd, Pb, and Ni [120], and also Co [121]. Fullerenes were found to exhibit better properties in metal preconcentration than more conventional sorbents such as ODS, activated carbon, and resins. C 60 fullerene was also tested as a sorbent for organic and organometallic compounds from aqueous solution [122]. For this purpose, C 6o -packed minicolums were inserted in continuous flow systems and gas chromatographic or flame atomic spectrometry was used for detection, depending on the nature of the compounds assayed. The fullerene was found to adsorb many types of organic substances (for example, N-methylcarbamates, phenols, PAHs, or amines), with efficiencies depending on the nature of the compound concerned. Nevertheless, conventional sorbents such as XAD-2 (a nonionic polystyrene divinylbenzene resin) or polyurethane foam were more efficient than C 60 for this purpose. As an illustration, the chromatographic areas for four phenolic compounds (phenol, 3,4-dimethylphenol, 2-tert~butylphenol and 4-chlorophenol) obtained with a variable pH are shown in Fig. 14.16, in which these results are also compared with those provided by XAD-2 as sorbent and ethyl acetate as eluent. The optimal pH range was wider with C 60 (1-9.5) than with XAD-2 (1-6.5). However, the signal obtained with C 60 was smaller, which was attributed [122] to its lower sorption capacity for these compounds. The adsorption efficiency of C 60 was observed to decrease with the increasing polarity of the organic compounds, which is consistent with the adsorption mechanism assigned to fullerenes via 1T-electron interactions. In addition, when using solutes that contain polar groups, dispersion prevails over retention. Organometallic compounds such as metalocenes and organoleads were quantitatively adsorbed on C 60 via the formation of neutral complexes or chelates, and the adsorption constant was dramatically increased by the use of classical reagents such as pyrrolidinedithiocarbamate or diethyldithiocarbamate. It was therefore concluded [122] that the fullerenes possess a high analytical potential for preconcentrating organometallic compounds, which is superior to that of conventional sorbents such as RP-C18, silica gel and activated carbon.
14.10
Adsorption from Solution: Colloidal and Biological Systems
~-~~_.-
M
-
o
..
_~
- - - -- -
~6
1
.. _- .. -
. - - - - :
~ : : - -)IE ............. 2 ...... "it.. . "'''''', ...... ,
as
:E
a.
3
•l
,,
-------1'--- --. ---.. "" . .
o
E
\,
4
C)
o
}fE
'"'.at-. .
~
Cti
*. "...
CD
(ij o
357
I
~
2
.c () 2
468
10
Sample pH
Figure 14.16 Influence of the sample pH on the adsorption of phenols on XAD-2 (dashed line) and C 60 (solid line). 1 = 2-tert-butylphenol; 2 = 3,4-dimethylphenol; 3 = 4-chlorophenol; 4 = phenol. Sample = 10 ml of aqueous solution containing 100 ng/ml of each phenolic compound. (Adapted from Re£ [122] with permission from Elsevier.)
Dithiocarbamate fungicides containing different metal ions such as Zn 2 +, Mn 2 + and Fe3+ were resolved using a C 60 column that performed better than the conventional RP-C18 sorbent [123]. However, the method did not allow the speciation of dithiocarbamates that contain no metal. Finally, C 6o -fullerene was reported to be an effective sorbent material for preconcentrating mercury compounds, and to be preferable to RP-C 18 on account of its large specific surface area and volume, which endow it with an increased physical sorption capacity [124]. Also recently, Pereira et al. [125] developed a new alternative for Cd and Pb determinations at low concentrations, using the preconcentration of a C 60 /C 70 mixture coupled to thermospray flame furnace atomic absorption spectrometry.
14.10 ADSORPTION FROM SOLUTION: COLLOIDAL AND BIOLOGICAL SYSTEMS In a review on the relationship of fullerenes with biological sciences, fullerenes and their derivatives were reported by Jensen et al. [126] to influence biological processes "in vivo." However, the mechanism was not yet fully understood as interactions of fullerenes at the biomolecular level had not been sufficiently studied. This was due, in turn, to the low solubility of fullerenes in water, a drawback which has led to some work on the production of stable water solutions of fullerenes. Indeed, the low solubility in water of fullerenes and the
358
Chapter 14 Adsorption on Fullerenes
limited accessibility of their inner spaces have significantly dampened the study of fullerene interactions with molecular species, and therefore alternatives such as water-soluble "nanographites" [127, 128] have been developed as a substitute for fullerenes in research being carried out in the interdisciplinary field between intermolecular and interfacial interactions. Several methods for producing stable aqueous dispersions of C 60 and C 70 without the addition of any stabilizer have been reported [129-131]. These methods are based on the exchange of molecules of an organic solvent, which covers the fullerenes, with water molecules. The resulting aqueous solutions of fullerenes, termed fullerene-water systems (FWSs), have the properties of colloidal systems. This type of dispersion has a high stability, no essential changes taking place during several months of storage at normal conditions [132-134]. The molecular system of C 60 in water (C 6o FWS) contains individual hydrated fullerenes and their fractal clusters in a hydrated state. The stabilization mechanism of such dispersions is not clear at the moment. Nevertheless, a negative charge of colloidal particles detected in different works [130, 131, 135, 136] seems to play a significant role in this stabilization. High resolution transmission electron microscopy and small-angle neutron scattering have revealed the polycrystalline nature of the clusters. Using a different approach, aqueous solutions of fullerenes were obtained by means of a chemical method based on oxidation of the C(;o anion in watermixable organic solvents [137]. The authors checked to see whether the sol was made up of hydrated C 60 . The colloidal particles were about 10 nm in diameter, and they were noncrystalline and quite homogeneous. Mchedlov-Petrossyan et al. [138] studied the interaction of cationic dyes such as indopolycarbocyanine and methylene blue with the C 6o FWSs, and demonstrated the occurrence of a strong interaction between the cationic dyes and the dispersed phase of the C 60 hydrosol, which resulted in adsorption at the surface of the colloid particles and finally in the coagulation of the sol [131, 134] Such adsorption processes are accompanied by the neutralization and hydrophobization of the C 60 /water interface and play a decisive role in coagulation phenomena. The same team [139] also studied the interaction of human serum albumin with hydrated fullerenes using electron spin resonance and differential scanning microcalorimetry. Their results suggested that the thermal stability of the protein, the surface tension of this protein-water matrix and the dynamics of water molecules in the vicinity of the protein surface are affected by hydrated fullerenes in water-salt solutions. The authors attributed these effects to similarities between protein and fullerene hydrations which cause long-range protein-protein, fullerene-fullerene, and protein-fullerene interaction forces and probably entropic depletion. They also suggested that the hydrated C 6o -induced stabilization of protein clusters can lead to the formation of polarized multilayers of water similar to those discovered in living cells, and that this can modify the biological activity of proteins and support the osmotic homeostasis of biological liquids.
References
359
14.11 CONCLUSIONS There is reasonable agreement among authors regarding the nature of the adsorption sites for gases and vapors on solid C 60 . At least three sites on the surface of C 60 have been identified, the energy decreasing in the following sequence: void space between four neighboring C 60 molecules > void space between two neighboring C 60 molecules > on top of a C 60 molecule. This is supported by studies with different adsorbates (N2 , Ar, CO 2 , CO, alkanes, alkenes) and techniques (isotherms, IGC, IR spectra, GCMC simulations). Concerning the possible use offullerenes for hydrogen storage, there is enough evidence to suggest that hydrofullerenes could be prepared to different degrees of hydrogenation. The easiest material to prepare would be C 6o H 36 , which has a gravimetric storage density of 4.5 wt% (close to DOE's requirements). However, there is a serious discrepancy concerning their dehydrogenation since some works suggest that the original fullerene could be recovered, whereas others indicate that dehydrogenation would be accompanied by irreversible fullerene decomposition. The efficiency and selectivity of fullerenes as adsorbents from aqueous solutions has resulted in a number of analytical applications of C 60 and C 70 as chromatography stationary phases, as chemical sensors and, especially, as sorbents for the preconcentration of analytes. In the latter case, the adsorption properties of fullerenes are more useful for inorganic and organometallic compounds than for organic compounds. On the other hand, the fullerenes exhibit a selectivity for aromatic compounds and planar molecules that makes them very attractive as stationary phases for liquid chromatography.
ACKNOWLEDGMENTS Financial support from the Spanish CSIC is gratefully acknowledged.
REFERENCES 1. Kroto, H.W., Heath, J.R., O'Brien, S.A., et al. (1985). C 60 Buckminsterfullerene. Nature, 318, 162-3. 2. Kratschmer, W., Lamb, L.D., Fostiropoulos, K., and Huffman, D.R. (1990) Solid C 60 - a new form of carbon. Nature, 347, 354-8. 3. Bandosz, T., Biggs, MJ., Gubbins, K.E., et al. (2003). Molecular models for porous carbons. Chern. Phys. Carbon, 28, 41-228.
360
Chapter 14 Adsorption on Fullerenes
4. Dresselhaus, M.S., Dresselhaus, G., and Eklund, P.C. (1996). Science of Fullerenes and Carbon Nanotubes. Academic Press. 5. Sastre, G., Cano, M.L., Corma, A., et al. (1997). On the incorporation of buckminsterfullerene C 60 in the supercages of zeolite Y. J. Phys. Chem. B, 101, 10184-90. 6. Zajac,]. and Groszek, AJ. (1997). Adsorption of C60 fullerene from its toluene solutions on active carbons: application of flow microcalorimetry. Carbon, 35, 1053-60. 7. Ajima, K., Yudasaka, M., Suenaga, K., et al. (2004). Material storage mechanism in porous nanocarbon. Adv. Mater., 16, 397-401. 8. Baletto, F., Doye, ].P.K., Ferrando, R., and Mottet, C. (2003). Adsorption and diffusion on nanoclusters of C 60 molecules. Surf. Sci., 532, 898-904. 9. Mackie, E.B., Galvan, D.H., and Migone, A.D. (2000). Methane adsorption on planar WS 2 sand on WS 2 -fullerene and -nanotube containing samples. Adsorption, 6, 169-74. 10. Chen,]. and Wu, F. (2004). Review of hydrogen storage in inorganic fullerenelike nanotubes. Appl. Phys. A, 78, 989-94. 11. Ismail, I.M.K. and Rodgers, S.L. (1992). Comparisons between fullerene and forms of well-known carbons. Carbon, 30, 229-39. 12. Cuesta, A.,]amond, M., Martinez-Alonso, A., and Tasc6n,].M.D. (1996). Thermal behavior of fullerenes in different atmospheres. Carbon, 34, 1239-48. 13. Kaneko, K., Ishii, C., Arai, T., and Suematsu, H. (1993). Defect-associated microporous nature of C60 crystals. J. Phys. Chem., 97, 6764-6. 14. Rostovtsev, R., Ishii, C., Setoyama, N., et al. (1996). Recrystallization-induced defect porous nature of C 60 crystals. Adsorption, 2, 153-6. 15. Rathousky, ]., Starek, ]., Zukal, A., and Kratschmer, W. (1993). Uptake of cyclopentane vapours in a fullerite powder. Fullerene Sci. Technol., 1, 575-82. 16. Kratschmer, W., Rathousky, ]., and Zukal, A. (1999). Adsorption of krypton at 77 K on fullerene C 60 , graphitized carbon black and diamond. Carbon, 37, 301-5. 17. Rathousky,]. and Zukal, A. (2000). Adsorption of krypton and cyclopentane on C 60 : an experimental study. Fullerene Sci. Technol., 8, 337-50. 18. Belz, T., Bauer, A., Find,]., et al. (1998). Structural and chemical characterization ofN-doped nanocarbons. Carbon, 36, 731-41. 19. Beck, M.T., Mandy, D., Papp, S., and Dekany, I. (2004). Surface porosity of fullerene black adsorbents modified by the Diels-Alder reaction. Carbon, 42, 677-9. 20. Beck, M.T., Mandy, D., Papp, S., and Dekany, 1.(2004). Surface modification of activated carbon and fullerene black by Diels-Alder reaction. Colloid Polym. Sci., 283, 237-42. 21. Cascarini de Torre, L.E., Fertitta, A.E., Flores, E.S., et al. (2004). Characterization of shungite by physical adsorption of gases. J. Argent. Chem. Soc., 92, 51-8. 22. Nagano, Y., Kiyobayashi, T., and Nitta, T. (1994). CO 2 absorption in C 60 solid. Chem. Phys. Lett., 217, 186-90. 23. Gusev, V.Y., Ruetsch, S., Popeko, L.A., and Popeko, ].E. (1999). Nitrogen and argon adsorption and SEM characterization of C 60 and C 60 / C 70 fullerenes: comparison with graphite. J. Phys. Chem. B, 103, 6498-503. 24. Martinez-Alonso, A., Tasc6n, ].M.D., and Bottani, EJ. (2000). Physisorption of simple gases on C 60 fullerene. Langmuir, 16, 1343-8.
References
25. Martinez-Alonso, A., Tasc6n, J.M.D., and Bottani, E.J. (2001). Physical adsorption of Ar and CO 2 on C 60 fullerene.]. Phys. Chern. B, 105, 135-9. 26. Tasc6n, J.M.D. and Bottani, E.J. (2002).Nitrogen physisorption on defective C 60 .]. Phys. Chern. B, 106,9522-7. 27. Tasc6n,].M.D. and Bottani, EJ. (2004). Ethylene physisorption on C 60 fullerene. Camon, 42, 1333-7. 28. Haghseresht, F., Finnerty, J.J., Nouri, S., and Lu, G.Q. (2000). Adsorption of aromatic compounds onto activated carbons: effects of the orientation of the adsorbates. Langmuir, 18, 6193-200. 29. Vanderbilt, D. and Tersof£J. (1992). Negative-curvature fullerene analog ofC 6o . Phys. Rev. Lett., 68, 511-13. 30. Jiang,J. and Sandler, S.1. (2003). Monte Carlo simulation of0 2 and N 2 adsorption in nanoporous carbon (C 168 schwarzite). Langmuir, 19,3512-18. 31. ]iang,]. and Sandler, S.1. (2003). Monte Carlo simulation of 02 and N 2 mixture adsorption in nanoporous carbon (C 168 schwarzite). Langmuir, 19, 5936-41. 32. Jiang,]., Klauda,].B., and Sandler, S.1. (2004). Hierarchical modeling O 2 and N 2 adsorption in C 168 schwarzite: from quantum mechanics to molecular simulation. ]. Phys. Chern. B, 108, 9852-60. 33. Jiang, J. and Sandler, S.1. (2005). Separation of CO 2 and N 2 by adsorption in C 168 schwarzite: a combination of quantum mechanics and molecular simulation study.]. Am. Chem. Soc., 127,11989-97. 34. Bernal, M.P., Coronas, J., Menendez, M., and Santamaria, J. (2004). Separation of CO 2/N 2 mixtures using MFI-type membranes. AIChE]., 50, 127-35. 35. Jiang,J., Klauda,J.B., and Sandler, S.1. (2005). Hierarchical modelling N 2 adsorption on the surface of and within a C 60 crystal: from quantum mechanics to molecular simulation.]. Phys. Chem. B, 109,4731-7. 36. Arora, G., Klauda,].B., and Sandler, S.1. (2005). A comparative study of nitrogen physisorption on different C 70 crystal structures using an ab initio based potential. ]. Phys. Chern. B, 109, 17267-73. 37. Oh, D.H. and Lee, Y.H. (1995). Orientational ordering of solid C 70 . Phys. Rev. Lett., 75, 4230-3. 38. Breton, ]., Gonzalez Platas, ]., and Girardet, C. (1993). Endohedral and exohedral adsorption in C 60 ; an analytical model.]. Chem. Phys., 99, 4036-40. 39. Iglesias-Groth, S., Breton, ]., and Girardet, C. (1998). Structure of the van der Waals rare gas-C 6o exohedral complexes [(C 6o )(RG)n; n == 1,2]. Chem. Phys., 237, 285-93. 40. Dawid, A. and Gburski, Z. (2003). Interaction-induced light scattering in a fullerene surrounded by an ultrathin argon "atmosphere": molecular dynamics simulation. Phys. Rev. A, 68, 065202. 41. Dendzik, Z., Kosmider, M., Dawid, A., et al. (2005). Interaction-induced depolarized light scattering spectra of exohedral complexes of N e and Ar with fullerenes and nanotubes. Mater. Sci.-Poland, 23, 457-66. 42. Szybisz, L. and Urrutia, I. (2004). Adsorption of 4He on a single C 60 .]. Low- Temp. Phys., 134, 1079-96. 43. Sartarelli, S.A., Szybisz, L., and Hernandez, E.S. (2005). 3He impurities in 4He systems adsorbed on curved substrates.]. Low- Temp. Phys., 138, 979-93. 44. Abraham, M.H., Du, C.M., Grate, ].W., et al. (1993). Fullerene as an adsorbent of gases and vapors.]. Chem. Soc. Chem. Commun., 1863-4.
Chapter 14 Adsorption on Fullerenes
45. Grate, J.W., Abraham, M.H., Du, C.M., et al. (1995). Examination of vapor sorption by fullerene, fullerene-coated surface acoustic wave sensonrs, graphite, and low-polarity polymers using linear solvation energy relationships. Langrnuir, 11, 2125-30. 46. Davydov, V.Ya., Kalashnikova, E.V., Karnatsevich, V.L., and Lopatin, M.A. (2000). Thermodynamic characteristics of adsorpion of organic compounds on molecular crystals of C60 fullerene. Russ.]. Phys. Chern., 74, 619-24. 47. Papirer, E., Brendle, E., Ozil, F., and Balard, H. (1999). Comparison of the surface properties of graphite, carbon black and fullerene samples, measured by inverse gas chromatography. Carbon, 37, 1265-74. 48. Bottani, E.J. and Tasc6n, J.M.D. (2004). Energetics of adsorption of gases and vapors on carbons. Chern. Phys. Carbon, 29, 209-423. 49. Chao, Y.-C. and Shih, J.-S. (1998). Adsorption study of organic molecules on fullerene with piezoelectric crystal detection system. Anal. Chirn. Acta, 374, 3946. 50. Hayashi, A., Yamamoto, S., Suzuki, K., and Matsuoka, T. (2004). The first application offullerene polymer-like materials, C 60 Pd n , as gas adsorbents.]. Mater. Chern., 14, 2633-7. 51. Kartsova, L.A. and Makarov, A.A. (2004). New fullerene-based stationary phases for gas chromatography.]. Anal. Chern., 59, 724-9. 52. Davydov, V.Y., Kalashnikova, E.V., Karnatsevich, V.L., and Kirillow, A.L. (2004). Adsorption properties of multi-wall carbon nanotubes. Fullerenes Nanotubes Carbon Nanostruct., 12, 513-18. 53. Mixteco-Sanchez, J.C. and Guirado-Lopez, R.A. (2003). Contrasting bonding behaviour of thiol molecules on carbon fullerene structures. Phys. Rev. A 68, 053204. 54. Pchelarov, G., Topalova, I., Koprinarov, N., et al. (1997). Investigation of sorption by an electrode deposit. Carbon, 35, 755-8. 55. Chen, C., Chen, J., Wang, X., et al. (2000). Fullerenes-extracted soot: a new adsorbent for collecting volatile organic compounds in ambient air.]. Chrornatogr. A, 886, 313-17. 56. Samonin, V.V. and Slutsker, E.M. (2005). The kinetics of adsorption of organic solvent vapors on fullerene blacks. Russ.]. Phys. Chern., 79, 87-90. 57. Samonin, V.V. and Slutsker, E.M. (2005). The ability of fullerene blacks to adsorb substances of various natures from the gas phase. Russ.]. Phys. Chern., 79, 91-6. 58. Vijayakrishnan, V., Santra, A.K., Pradeep, T., et al. (1992). Interaction ofnitrogen and oxygen with C 60 .]. Chern. Soc., Chern. Cornrnun., 198-9. 59. Vijayakrishnan, V., Santra, A.K., Pradeep, T., et al. (1992). A photoelectron spectroscopic study of the interaction of oxygen, nitrogen and transition metals with C 60 films. Indian]. Chern., 31A-B, F22-F26. 60. Werner, H., Schedel-Niedrig, Th., Wohlers, M., et al. (1994). Reaction of molecular oxygen with C60: spectroscopic studies.]. Chern. Soc., Faraday Trans., 90, 403-9. 61. Wohlers, M., Werner, H., Belz, T., et al. (1997). C 60 : a host lattice for the intercalation of oxygen? Mikrochirn. Acta, 125, 401-6. 62. WU, J.X., Liu, X.M., Ma, M.S., et al. (2000). Photoemission study of oxygen adsorption on Rb 6 C 6o film surface. Appl. Surf. Sci., 153, 150-5.
References
63. Tanaka, K., Choo, C.-K., Sumi, S., et al. (2002). C 6o /zeolite semiconductor electrode and the gas sensing.]. Phys. Chern. B, 106,4155-61. 64. Matsumoto, K., Maeda, H., Postava, K., et al. (2003). Spectro-ellipsometric characterization and gaseous occlusion of fullerene C 60 crystals. Fullerenes Nanotubes Carbon Nanostruct., 11, 15-23. 65. Niklowitz, P.G., Li, Z.Y.,Jardine, A.P., et al. (2004). Physosorption ofmolecular oxygen on C60 thin films.]. Chern. Phys., 120, 10225-30. 66. Fastow, M., Kozirovski, Y., Folman, M., and Heidberg, J. (1992). IR spectra of CO and NO adsorbed on C 60 . ] . Phys. Chern., 96, 6126-8. 67. Fastow, M., Kozirovski, Y., and Folman, M. (1993). IR spectra of CO 2 and N 2 0 adsorbed on C 60 and other carbon allotropes - a comparative study.]. Electron Spectrosc. Relat. Phenorn., 64-65, 843-8. 68. Heidberg,J., Elstner,J., Lassmann, W., and Folman, M. (1993). Polarized FTIRspectra of C 60 layers and the adssorbates CO and NO on C 60 .]. Electron Spectrosc. Relat. Phenom., 64/65, 883-92. 69. Fastow, M., Kozirovski, Y., and Folman, M. (1995). Adsorption potentials and spectral shifts of CO adsorbed on C 60 . Surf. Sci., 331-333, 121-6. 70. Lubezky, A., Chechelnitsky, L., and Folman, M. (1996). IR spectra of CH 4 , CD 4 , C 2 H 4 , C 2 H 2 , CH3 0H and CH 3 0D adsorbed on C 60 films.]. Chem. Soc., Faraday Trans., 92, 2269-74. 71. Lubezky, A., Chechelnitsky, L., and Folman, M. (1996) IR spectra of CO adsorbed on mixed films of LiFjC 60 and NaCljC 6o . Surf. Sci., 368, 342-7. 72. Folman, M., Fastow, M., and Kozirovski. (1997). Surface heterogeneity of C 60 as studied by infrared spectroscopy of adsorbed CO and adsorption potential calculations. Langmuir, 13, 1118-22. 73. Dillon, A.C., Jones, K.M., Bekkedahl, T.A., et al. (1997). Storage hydrogen in single-walled carbon nanotubes. Nature, 386, 377-9. 74. Chambers, A., Park, C., Baker, R.T.K., and Rodriguez, N.M. (1998). Hydrogen storage in graphite nanofibers.]. Phys. Chem. B., 102, 4253-6. 75. Strobel, R., Jorissen, L., Schliermann, T., et al. (1999). Hydrogen adsorption on carbon materials. J. Power Sourc., 84, 221-4. 76. Tibbetts, G.G., Meisner, G.P., and Olk, C.H. (2001). Hydrogen storage capacity of carbon nanotubes, filaments, and vapor-grown fibers. Carbon, 39, 2291-301. 77. Lee, S.M., An, K.H., Kim, W.S., et al. (2001). Hydrogen storage in carbon nanotubes. Synth. Met., 121, 1189-90. 78. Dillon, A.C., and Heben, MJ. (2001). Hydrogen storage using carbon adsorbents: past, present and future. Appl. Phys. A: Mater. Sci. Process., 72, 133-42. 79. Ziittel, A. and Orimo, S. (2002). Hydrogen in nanostructured, carbon-related, and metallic materials. MRS Bull., 27, 705-11. 80. Schur, D.V., Tarasov, B.P., Shul'ga, Y.M., et al. (2003). Hydrogen in fullerites. Camon, 41, 1331-42. 81. Haufler, R.E., Conceicao,J., Chibante, L.P.F., et al. (1990). Efficient production of C 60 (buckminsterfullerene), C 6o H 36 , and the solvated buckide ion. J. Phys. Chem., 94, 8634-6. 82. Kolesnikov, A.I., Antonov, V.E., Bashkin, 1.0., et al. (1999). Neuton spectroscopy of fullerite hydrogenated under high pressures. Physica B, 263-264, 436-8. 83. Ye, Y., Ahn, C.C., and Fultz, B. (2000). Hydrogen adsorption and phase transitions in fullerite. App. Phys. Lett., 77, 2171-3.
Chapter 14 Adsorption on Fullerenes
84. Jin, C., Hettich, R., Compton, R., et al. (1994). Directe solid-phase hydrogenation offullerenes.J. Phys. Chem., 98, 4215-17. 85. Meletov, K.P., Assimopoulos, S., Tsilika, I., et al. (2001). Isotopic and isomeric effects in high-pressure hydrogenated fullerenes studied by Raman spectroscopy. Chem. Phys., 263, 379-88. 86. Kurmaev, E.Z., Moewes, A., Ida, T., et al. (2003). Isomer structure of highpressure hydrofullerene probed by soft X-ray emission. J. Mol. Struct. (Theochem), 639,27-33. 87. Tarasov, B.P., Fokin, V.N., Moravsky, A.P., et al. (1997). Hydrogenation of fullerenes C 60 and C 70 in the presence of hydride-forming metals and intermetallic compounds. J. Alloys Compd., 253-254, 25-8. 88. Brosha, E.L., Davey, J., Garzon, F.H., and Gottesfeld, S. (1999). Irreversible hydrogenation of solid C 60 with and without catalytic metals. J. Mater. Res., 14, 2138-46. 89. Bini, R. Ebenhoch, J., Fanti, M., et al. (1998). The vibrational spectroscopy of C 6o H 36 : an experimental and theorical study. Chem. Phys., 232, 75-94. 90. Bulusheva, L.G., Okotrub, A.V., Antich, A.V., and Lobach, A.S. (2001). Ab initio calculation of X-ray emission and IR spectra of the hydrofullerene C 6o H 36 . J. Mol. Struct., 562, 119-27. 91. Dorozhko, P.A., Lobach, A.S., Popov, A.A., et al. (2001). Sublimation of hydrofullerenes C 6o H 36 and C 6o H 18 . Chem. Phys. Lett., 336, 39-46. 92. Bensasson, R.V., Hill, T.J., Land, E.J., et al. (1997). Spectroscopy and photophysics ofC 60 H 18 and C 6o H 36 . Chem. Phys., 215,111-23. 93. Gakh, A.A., Romanovich, Y.A., and Bax, A. Thermodynamic rearrangement synthesis and NMR structures of C 1 , C 3 and T isomers of C 6o H 36 . J. Am. Chem. Soc., 125, 7902-6. 94. Tiirker, L. and Erkoc;, S. (2003). AMl treatment of endohedrally hydrogen doped fullerene, nH 2 @C 60 . J. Mol. Struct. (Theochem) , 638, 37-40. 95. Narita, I. and Oku, T. (2002). Molecular dynamics calculation of H 2 gas storage in C 60 and B36 N 36 clusters. Diam. Relat. Mater., 11, 945-8. 96. Barajas-Barraza, R.E. and Guirado-L6pez, R.A. (2002). Clustering of H 2 molecules encapsulated in fullerene structures. Phys. Rev. B., 66, 155426. 97. Tiirker, L. (2003). AMl treatment of (Li + nH 2 )n=o-S@C 60 systems. Int. J. Hydrogen Energy, 28, 223-8. 98. Oskengorn, B. (2003). Preparation du complexe endoedrique hydrogene moleculaire fullerene C 60 , associe a du C 60 hydrogene. C. R. Chim., 6, 467-72. 99. Turnbull, J.D. and Boninsegni, M. (2005). Adsorption of para-hydrogen on fullerenes. Phys. Rev. B, 71, 205421. 100. Berezkin, V.I., Viktorovski, LV., Vul, A.Ya., et al. (2002). A comparative study of the sorption capacity ofactivated charcoal, soot, and fullerenes for organochlorine compounds. Tech. Phys. Lett., 28, 885-8. 101. Berezkin, V.I., Viktorovski, LV., Vul, A.Ya., et al. (2003). Fullerene single crystals as adsorbents of organic compounds. Semiconductors, 37, 775-83. 102. Cheng, X., Kan, A.T., and Tomson, M.B. (2004). Naphthalene adsorption and desorption from aqueous C 60 fullerene. J. Chem. Eng. Data, 49, 675-83. 103. Cheng, X.K., Kan, A.T., and Tomson, M.B. (2005). Uptake and sequestration of naphthalene and 1,2-dichlorobenzene by C 60 .J. Nanoparticle Res., 7, 555-67. 104. Baena, J.R., Gallego, M., and Valcarcel, M. (2002). Fullerenes in the analytical sciences. Trends Anal. Chem., 21, 187-98.
References
105. Saito, Y., Ohta, H., Terasaki, H., et al. (1995). Separation of polycyclic aromatic hydrocarbons with a C 60 bonded silica phase in microcolumn liquid chromatography.]. High-Res. Chromatogr., 18, 569-72. 106. Jinno, K., Tanabe, K., Saito, Y., and Nagashima, H. (1997). Separation of polycyclic aromatic hydrocarbons with various C 60 fullerene bonded silica phases in microcolumn liquid chromatography. Analyst, 122, 787-91. 107. Stalling, D.L., Guo, C.Y., and Saim, S. (1993). Surface-linked-C(60/70)polystyrene divinylbenzene beads as a new chromatographic material for enrichment of coplanar PCBS.]. Chromatogr. Sci., 31, 265-78. 108. Bianco, A., Gasparrini, F., Maggini, M., et al. (1997). Molecular recognition by a silica-bound fullerene derivative.]. Am. Chem. Soc., 119, 7550-4. 109. Golovnya, R.V., Terenina, M.B., Ruchkina, E.L., and Karnatsevich, V.L. (1993). Fullerene C 60 as a new stationary-phase in capillary gas-chromatography. Mendeleev Commun., 119, 231-3. 110. Glausch, A., Hirsch, A., Lamparth, I., and Schurig, V. (1998). Retention behaviour of polychlorinated biphenyls on polysiloxane-anchored C 60 in gas chromatography.]. Chromatogr. A, 809, 252-7. 111. Chen, Y.Y., Fang, P.F., Zeng, Z.R., and Fan,J.H. (1999). Synthesis of linear fullerene-containing polysiloxanes and their application to capillary gas chromatography. Chem. Lett., 28, 499-500. 112. Fang, P.-F., Zeng, Z.-R., Fan, J.-H., and Chen, Y.-Y. (2000). Synthesis and characteristics of [60]fullerene polysiloxane stationary phase for capillary gas chromatography.]. Chromatogr. A, 867, 177-85. 113. Shih, J.-S., Chao, Y.-C., Sung, M.-F., et al. (2001). Piezoelectric crystal membrane chemical sensors based on fullerene C60. Sense Actuators B, 76, 347-53. 114. Wang, L.-G., Wang, X., Ottova, A.L., and Tien, H.T. (2005). Iodide sensitive sensor based on a supported bilayer lipid membrane containing a cluster form of carbon (fullerene C60). Electroanalysis, 8, 1020-2. 115. Gavalas, V.G. and Chaniotakis, N.A. (1998). [60]Fullerene mediated amperometric biosensors. Anal. Chim. Acta, 409, 131-5. 116. Amao, Y., Asai, K., and Okura, I. (2000). A novel optical oxygen sensing system based on triplet-triplet reflectance of fullerene C 6o -polystyrene film by time-resolved spectroscopy using diffuse reflectance laser flash photolysis. Analyst, 125,523-6. 117. Gallego, M., Petit de Pefia, Y., and Valcarcel, M. (1994). Fullerenes as sorbent materials for metal preconcentration. Anal. Chem., 66, 4074-8. 118. Petit de Pefia, Y., Gallego, M., and Valcarcel, M. (1997). Fullerene: a sensitive and selective sorbent for the continuous preconcentration and atomic absorption determination of cadmium.]. Anal. Atom. Spectrosc., 12, 453-7. 119. Petit de Pefia, Y., Gallego, M., and Valcarcel, M. (1995). Preconcentration of copper traces on C 60 -C70 fullerenes by formation of ion pairs and chelates. Anal. Chem., 67, 2524-9. 120. Silva, M.M., Arruda, M.A.Z., Krug, F.J., et al. (1998). On-line separation and preconcentration of cadmium, lead and nickel in a fullerene (C 60 ) minicolumn coupled to flow injection tungsten coil atomic absorption spectrometry. Anal. Chim. Acta, 368, 255-63. 121. Gonzalez, M.M., Gallego, M., and Valcarcel, M. (1999). Effectiveness offullerene as a sorbent for the determination of trace amounts of cobalt in wheat flour
366
122.
123. 124.
125.
126. 127.
128.
129.
130.
131.
132.
133.
134.
135.
136.
Chapter 14 Adsorption on Fullerenes
by electrothermal atomic absorption spectrometry. J. Anal. Atom. Spectrosc., 14, 711-6. Ballesteros, E., Gallego, M., and Valcarcel, M. (2000). Analytical potential of fullerene as adsorbent for organic and organometallic compounds from aqueous solutions. J. Chromatogr. A, 869, 101-10. Baena, J.R., Gallego, M., and Valcarcel, M. (2000). Group speciation of metal dithiocarbamates by sorption on C 60 fullerene. Analyst, 125, 1495-9. Munoz, J., Gallego, M., and Valcarcel, M. (2004). Solid-phase extraction-gas chromatography-mass spectrometry using a fullerene sorbent for the determination of inorganic mercury(II), methylmercury(I) and ethylmercury(I) in surface waters at sub-ng/m1levels. J. Chromatogr. A, 1055, 185-90. Pereira, M.G., Pereira Filho, E.R., Berodt, H., and Arruda, M.A.Z. (2004). Determination of cadmium and lead at low levels by using preconcentration at fullerene coupled to thermospray flame furnace atomoc absorption spectrometry. Spectrochim. Acta B, 59, 515-21. Jensen, A.W., Wilson, S.R., and Schuster, D.1. (1996). Biological applications of fullerenes. Biootg. Med. Chem., 4, 767-79. Kamegawa, K., Nishikubo, K., Kodama, M., et al. (2003). Dissolutionaggregation behaviour of water-soluble nanographites and their adsorptive characteristics for 2-naphtol in aqueous solutions. J. Colloid Interface Sci., 268, 58-62. Kamegawa, K., Nishikubo, K., Kodama, M., et al. (2005). Aqueous-phase adsorption of aromatic compounds on water-soluble nanographite. Colloids Surf., 254, 31-5. Andrievsky, G.V., Kosevich, M.V., Vovk, O.M., et al. (1995). On the production of an aqueous colloidal solution of fullerenes. J. Chem. Soc., Chem. Commun., 12, 1281-2. Deguchi, S., Alargova, R.G., and Tsujii, K. (2001). Stable dispersions of fullerenes, C 60 and C 70 , in water. Preparation and characterization. Langmuir, 17, 6013-7. Mcheldov-Petrossyan, N.O., Klochkov, V.K., and Andrievsky, G.V. (1997). Colloidal dispersions of fullerene C 60 in water: some properties and regularities of coagulation by electrolytes. J. Chem. Soc., Faraday Trans., 93, 4343-6. Prilutski, Yu.l., Durov, S.S., Yashchuk, V.N., et al. (1999). Theoretical predictions and experimental studies of self-organized C 60 nanoparticles in water solution and on the support. Bur. Phys.J. D, 9, 341-3. Andrievsky, G.V., Klochkov, V.K., Karyakina, E.L., and Mcheldov-Petrossyan, N.O. (1999). Studies of aqueous colloidal solutions of fullerene C 60 by electron microscopy. Chem. Phys. Lett., 300, 392-6. Mchedlov-Petrossyan, N.O., Klochkov, V.K., Andrievsky, G.V., et al. (1999). Interaction between cationic dyes and colloidal particles in a C 60 hydrosol. Mendeleev Commun., 2, 63-4. Andrievsky, G. V., Klochkov, V. K., Bordyuh, A., and Dovbeshko, G. I. (2002). Comparative analysis of two aqueous-colloidal solutions of C 60 fullerene with help ofFTIR reflectance and UV-Vis spectroscopy. Chem. Phys. Lett., 364, 8-17. Avdeev, M.V., Khokhryakov, A.A., Tropin, T.V., et al. (2004). Structural features of molecular-colloidal solutions of C 60 fullerenes in water by small-angle neutron scattering. Langmuir, 20, 4363-8.
References
137. Wei, X., Wu, M., Qi, L., and Xu, Z. (1997). Selective solution-phase generation of C 60 n- (n = 1, 2) and formation of an aqueous colloidal solution of C60. ]. Chern. Soc., Perkin Trans. 2, 1389-93. 138. Mcheldov-Petrossyan, N.G., Klochkov, V.K., Andrievsky, G.V., and Ishchenko, A.A. (2001). Interaction between colloidal particles of C 60 hydrosol and cationic dyes. Chern. Phys. Lett., 341, 237-44. 139. Rozhkov, S.P., Goryunov, A.S., Sukhanova, G.A., et al. (2003). Protein interaction with hydrated C 60 fullerene in aqueous solutions. Biochem. Biophys. Res. Commun., 303, 562-6.
HYDROGEN ADSORPTION IN SINGLE-WALLED CARBON NANOTUBES J. Karl Johnson 1,2 and Milton W. Cole 3 1 Department of Chemical & Petroleum Engineering, University of Pittsburgh, Pittsburgh, PA, USA 2National Energy Technology Laboratory, US Department of Energy, Pittsburgh, PA, USA 3Department of Physics, Pennsylvania State University, University Park, PA, USA
Contents 15.1 Introduction 15.2 Experiment, Simulation, and Theory of Hydrogen Storage 15.3 Quantum Sieving 15.4 Phase Transition Phenomena 15.5 Summary and Conclusions Acknowledgments References
36 9 370 385
39 1 393 393 394
15.1 INTRODUCTION This chapter is concerned with the properties of hydrogen molecules adsorbed within bundles of carbon nanotubes. A body of background information relevant to this problem is discussed in Chapter 9, which is concerned primarily with gases other than hydrogen in nanotubes. What is so unusual about hydrogen that justifies a separate discussion? The answer is the same as the reason why this particular gas has probably received more attention than all other gases combined. It is because many researchers have been investigating the possibility Adsorption by Carbons ISBN: 978-0-08-044464-2
© 2008 Elsevier Ltd. All rights reserved.
369
370
Chapter 15 Hydrogen Adsorption in Single-Walled Carbon Nanotubes
of storing significant quantities of hydrogen in nanotubes. This search was stimulated in large part by an experiment reported in Nature claiming to observe 5-10 wt% of hydrogen stored on single-walled carbon nanotubes (SWNTs) at room temperature and low pressure [1]. Such a high storage capacity, if confirmed, might well provide the basis for a vehicular fuel cell hydrogen storage technology. About the same time another experimental paper appeared, widely reported in the popular media, claiming to find more than 50 wt% hydrogen in carbon nanofibers [2, 3]. These "fantastic" results have never been confirmed by independent groups [4]. However, these reports were sufficient to start a flurry of experimental, theoretical, and simulation work on hydrogen adsorption on carbon nanotubes. This article describes three research topics in the subsequent sections. Section 15.2 describes experimental and theoretical research pertinent to the exploration of hydrogen storage capacity, mostly at room temperature. Section 15.3 discusses the problem of quantum sieving, which is the separation oflight isotopes, e.g., hydrogen from deuterium, by adsorption within nanotube bundles. Section 15.4 summarizes a variety of open questions concerning phase transition phenomena that have been proposed to occur for hydrogen within nanotube bundles.
15.2 ExPERIMENT, SIMULATION, AND THEORY OF HYDROGEN STORAGE
In this section we first briefly review the main experimental findings for hydrogen adsorption on SWNTs and then review in detail the simulation and theory findings. We focus mainly on the modeling of hydrogen adsorption in SWNTs because the experiments are so far hampered by a lack of pure and well-characterized nanotube samples. In many ways, experimental work is more of a materials issue, since the presence of catalyst particles, amorphous and graphitic carbon impurities, and chemical and geometrical defects on the nanotube samples make it difficult to unambiguously compare different experiments and interpret the observations. Experimental work on hydrogen adsorption in carbon nanotubes has yielded a wide range of sometimes conflicting results and several reviews have appeared [5-11]. Dillon and coworkers [1] were the first to report very high adsorption at room temperature, claiming to observe 5-10wt%. Chen et al. [12] observed apparent hydrogen adsorption of up to 20 wt% on alkali-doped SWNTs. These results were later shown to be due to water impurities reacting with the alkali metals [13, 14]. Liu and coworkers have measured reversible hydrogen adsorption of about 3-4 wt% on SWNTs at room temperature and pressures of about 10 Mpa [15, 16]. In contrast, there have been a number of publications from different groups finding much more modest hydrogen uptake by SWNTs at room temperature and pressures less than 30 Mpa
15.2
Experiment, Simulation, and Theory of Hydrogen Storage
371
[17- 28]. There are other experiments showing high levels of hydrogen adsorption on SWNTs at cryogenic temperatures, typically around 77 K [4, 29-31], but these conditions are not particularly relevant to vehicular fuel cells. One of the challenges of experimental measurements is that the uptake of hydrogen has been shown to be sensitive to the pretreatment of the SWNT samples [7, 25, 29, 32-34]. This challenge may also be an opportunity, as we will discuss later in this chapter. The simulations and theory of hydrogen adsorption on carbon nanotubes can be placed into three broad categories: (1) Modeling of physisorption using classical potentials. (2) Ab initio modeling of physisorption energies and geometries of molecular hydrogen on nanotubes. (3) Ab initio modeling of chemisorption of molecular and atomic hydrogen on nanotubes. We discuss each of these areas below. 15.2.1 Modeling of Physisorption with Classical Potentials
Statistical mechanical modeling with classical potentials has been used successfully to model physisorption of many different gases on microporous sorbents such as activated carbons and zeolites [35-43]. It is fairly common to observe quantitative agreement between experimentally measured quantities, such as isotherms, isosteric heats of adsorption, layering transitions, and monolayer ordering, and these same quantities computed from molecular simulations. The statistical mechanical calculations are essentially exact, to within statistical accuracy. However, there are two critical problems with these simulations. The first is that the potential models for both fluid-solid and fluid-fluid interactions are not known to good accuracy. The second is that the molecular level structure of the sorbent is not always 'known. Errors in the potential models or the sorbent geometry can lead to quantitative and even qualitative disagreement with experiments. These two problems are relevant issues with regard to modeling of hydrogen adsorption in carbon nanotubes. It is commonly assumed that the potential for H 2-nanotube interactions can be taken as the H 2 -graphene interactions with the graphene sheet wrapped into the appropriate nanotube. This is an untested hypothesis, since an unambiguous answer would require very high-level ab initio calculations, such as at the coupled cluster level of theory. These calculations are unfeasible with current algorithms and computer resources. We note that the van der Waals part of the Hz/Nanotube interaction requires knowledge of the anisotropic dielectric function of the tubes, as was done for Hz/graphite; see Vidali et al. [44] The analogous dielectric response problem for the nanotubes has not been solved, even in principle. A precise description of the SWNT sorbent is also problematic. In theory, nanotubes should form perfectly ordered hexagonal bundles, giving a structure as well-defined as zeolites. In practice, nanotubes contain significant quantities of metal catalyst particles, amorphous carbon impurities, and geometric and chemical defects in the nanotubes themselves. Thus, the accurate modeling of gas adsorption on SWNTs is a challenge at all levels. Nevertheless, statistical
Chapter 15 Hydrogen Adsorption in Single-Walled Carbon Nanotubes
37 2
modeling can yield important information about hypothetical limits of hydrogen storage and optimum geometric arrangements of nanotubes. There have been a number of simulations and theoretical calculations carried out that assess the hydrogen storage capacity of SWNTs [25, 45-63]. The first molecular simulations of hydrogen adsorption in carbon nanotubes were carried out by Darkrim and Levesque [45]. They simulated a square array of nanotubes, although nanotubes are known to form hexagonal arrays. The temperature and pressure studied corresponded to 293 K and 10 MPa. They varied the lattice spacing and nanotube diameter. They found optimal uptake in the case of a nanotube with a diameter of 11.75 A and a tube-tube distance (van der Waals gap) of 7 A. This is similar to optimum conditions found for hexagonal arrays from simulations by other groups. For example, Wang and Johnson [49] performed simulations at 298 K and 50 atm with arrays consisting of either (9, 9), or (12, \2), or (18, 18) nanotubes and found an optimum with a (9, 9) (diameter of 12.2A) array with a van der Waals gap of 6A. We note that the optimum was defined in terms of the excess adsorption, (15.1) where fads is the total amount adsorbed (assuming only pore volume in the sorbent), Pbulk is the bulk gas density, and Vfree is the free volume, or the volume of the system minus the volume occupied by the solid sorbent. In a molecular simulation the total amount adsorbed is computed directly. Computing the excess adsorption is ambiguous because there can be different definitions of the free volume, with particularly large consequences in a nanoscale geometry [64]. Rzepka et al. [47] compared adsorption in graphene slit pores with adsorption inside a nanotube, where the slit pore width was the same as the diameter of the nanotube. They found that in general the slit pore geometry was better for hydrogen uptake than the nanotube at most temperatures and pressures studied. The amount of hydrogen stored at room temperature was found to be small, on the order of 1 wt% at 10 MPa. These same general conclusions were later confirmed by more elaborate calculations of Wang and Johnson, who performed path integral simulations with accurate classical potentials to account for quantum effects in the translational motion ofH 2 • Quantum effects are fairly small for adsorption in pores that are more than two molecular diameters wide at temperatures greater than 77 K. For example, the volumetric adsorption at 77 K in a slit pore 10 A wide from the classical Rzepka and coworkers is only a few percent larger than the path integral values from Wang and Johnson at the same conditions. Levesque et al. [60] state that quantum effects reduce adsorption by 4% at 293 K and 20% at 77 K. Many different groups have concluded that slit pores are overall a better geometry for hydrogen storage, but that pure graphene structures, whether nanotubes or slit pores, were incapable of storing more than about 1 wt% at room temperature and 10 MPa. This is far short of the Department of Energy (DOE) hydrogen storage gravimetric target of around 6 wt% (which takes into account the weight of the tank and associated hardware) [65]. I".J
I".J
15.2
Experiment, Simulation, and Theory of Hydrogen Storage
373
1.8
o .~ 1.6 ~
"u «S
~1.4
()
Q)
::c
rn
1.2
::> 1.0
20
40 60 Storage pressure (atm)
80
100
Figure 15.1 The usable capacity ratio for hydrogen adsorbed in idealized slit pores and idealized nanotube arrays. The discharge pressure is 0.1013 MPa and the temperature is 298 K. The open circles, open diamonds, and open squares are data for slit pores with widths of6.15, 9.23, and 20.51 A, respectively. The filled triangles and filled circles are data for arrays of (18, 18) and (9, 9) SWNTs, respectively. (Reprinted with permission from Ref. [48] Copyright 1999 by the American Institute of Physics.)
One useful measure of the effectiveness of a sorbent for gas storage is the so-called usable capacity ratio (VCR). This is defined as the mass of the available fuel in a sorbent-Ioaded storage tank divided by the mass of available fuel in a compressed gas tank of the same size. Thus, the VCR is a measure of the effectiveness of adsorption compared with compressed gas storage at the same conditions. If the UCR value is less than unity then compressed gas is more effective than adsorption. If the UCR value is greater than unity then the sorbent system gives more deliverable hydrogen than compression. Figure 15.1 is a plot of the VCRs for idealized slit pores and idealized arrays of nanotubes. We note that the (9, 9) nanotube array is always less effective than simple gas compression, whereas the larger diameter nanotubes have VCRs that are above unity, although they are always
==-qM
(19.4)
P, T,/L
Relationship (19.4), known as Lippmann equation, permits the evaluation of the excess of charge at the electrode surface from the electrocapillary curve y == y(4)). For interfaces relatively simple such as the mercury/1 M aqueous KCl interface, Eqn (19.4) results in a parabola with a maximum at ay/a4> == 0, i.e., for null charge at the electrode surface. This condition corresponds to the potential of zero charge (Epzc ) for the electrode in the electrolyte solution. The second derivative of Eqn (19.4) represents the capacitance C
2y ) a ( a4>2
p. T.IL
(aqM) = - a4> p. T.IL = - C
(19.5)
When y == y( 4» is a perfect parabolic function, it results that d y / d4> is proportional to 4>, and the electrochemical interface is characterized by a constant value of C. For real systems, /a4> changes with 4>, and therefore a differential capacitance (Cd) has to be defined
aqM
qM
d Cd = ( dE
)
(19.6)
p. T.IL
where E is the electric potential difference across the capacitance. Values of r i and y can be obtained by integration of the electrocapillary curve provided that the value of E pzc is known. 19.1.2
Adsorption at Electrodes
In the absence of chemical or electrochemical processes, the adsorption of molecules, ions or both at the electrode surface becomes possible. This fact involves electrode-solvent, electrode-ionic species, and solvent-ionic species as the most relevant interactions and the possible contribution oflateral interactions. These interactions play an important role in the behavior of C vs E curves. The adsorption of either ions or neutral molecules on the electrode surface depends on i.e., on the applied electric potential. Correspondingly, the electric field at the electrochemical interface is an additional free-energy contribution that either favors or restricts the adsorption of species on the electrode from the ionic conducting phase. A variety of adsorption isotherms has been proposed to account for the behavior of different electrochemical systems. Among them are the Langmuir, Frumkin, and Temkin isotherms [2]. Frumkin and Temkin isotherms, at variance with the Langmuir one, include effects such as adsorbate-adsorbate or adsorbate-surface interactions.
qM'
482
Chapter 19 Electrochemical Behavior of Carbon Materials
Langmuir Q) C)
Q) C)
~
~
Q)
>
0
()
Q)
> 0
Temkin
()
Q) ()
Q) ()
«S
«S
't:
't:
CJ)
CJ)
'0
'0
::::J
::::J
Q) Q)
Q)
~
0,
C) Q)
Q)
0
0
log pressure
Potential difference
Figure 19.1 Comparison of chemical and electrochemical isotherms. The arrow indicates the shift of the curves as the lateral interactions term increases.
A comparison of the dependence of OJ, the degree of surface coverage by species i, on either P (chemical adsorption) or E (electrochemical adsorption) is shown in Fig. 19.1. At the electrochemical interface, adsorption of either charged or neutral molecules and charge transfer processes may occur simultaneously. Electroadsorption and electrodesorption processes play a key role in electrocatalytic reactions [2].
19.1.3 Relevant Kinetic Parameters The rate of an electrochemical reaction involving reactant i, expressed as dNj / dt, where N j is the number of moles of i electrolyzed at time t, is proportional to the faradaic current (I) flowing across the cell. However, as the electrode process is a heterogeneous reaction, its rate is usually expressed as moles s cm2 1 dNj
j
A dt
zjF
(19.7)
where A is the electrode area and j is the current density, i.e., j = II A. A basic problem in electrochemical kinetics is to determine the current (I) as a function of the applied potential (E), particularly under steady-state conditions. The departure of the electrode potential from the equilibrium value (Erev = N ernst potential) is the electrode polarization that is measured by the overpotential (1])
1] = E - Erev
(19.8)
19.1
A Brief Summary of Electrochemical Concepts
The overall current efficiency for the nth process is given by the ratio between the fraction of the number of coulombs (Qn) involved in the nth process and the total charge (Qr) passed across the cell (19.9) For a single electrochemical process p = 1. Generally, the rate of the electrode process is influenced by the mass transport of reactants to and products from the electrode surface, the proper electron transfer process, and the chemical reactions preceding or following the electron transfer. Accordingly, the value of YJ may involve a concentration (mass transport), activation (electron transfer), and ohmic (ohmic resistance) polarization contribution. Let us consider a simple redox reaction involving species 0 and R in the solution
(19.10)
kf and kb being the rate constants for the forward (£ cathodic) and backward (b, anodic) reactions. The net current flowing through the electrochemical interface is the algebraic sum of the currents If and I b of the partial reactions (19.11) Co (O,t) and CR (O,t) being the concentration of 0 and R on the electrode surface at time t. The rate constants depend on the overpotential
kf
= k°exp [aZF - R T (E = E 0)]
kb=k°exp [(l-a)ZF( RT E=E 0)]
(19.12)
(19.13)
kO being the standard rate constant, a the transfer coefficient assisting the reaction in the forward direction, and EO the standard potential of the redox reaction. The value of kO is related to the exchange current density (jo) of the reaction at the reversible potential. Equations (19.12) and (19.13) are usually expressed as Tafel relationships. For the cathodic reaction, the Tafel equation is
YJ=a+blnj
(19.14)
with a = RTjjoa and b = -RTjaF. A similar Tafel equation can be written for the anodic reaction with a = RTjjo(l- a) and b = RTj(l- a)F.
Chapter 19 Electrochemical Behavior of Carbon Materials
Most electrochemical processes can be described by complex reaction mechanisms with a rate-determining step (rds). Besides, a stoichiometric number of the rds is defined as the number of times the rds has to occur for every complete act of the overall reaction. From the temperature dependence of Eqs (19.12) and (19.13), the activation energy of the cathodic and anodic reactions at different values of 1] can be obtained.
19.2 THERMODYNAMIC DATA FOR CARBON ELECTRODES
Standard aqueous electrode potentials for reactions involving carbon have been calculated from the free energy of formation of carbon-containing compounds at different pH and temperature[3-6]. These data, displayed as potentialpH equilibrium diagrams, determine the domains of relative predominance of carbon as such or under a dissolved carbon-containing species such as methanol, aldehyde, acetic acid, carbonate, bicarbonate, or gaseous species such as methane, carbon dioxide, and carbon monoxide. As an example, a scheme of a typical E/pH equilibrium diagram for graphite/water at 25°C is shown in Fig. 19.2. Lines (a) and (b) represent Nernst equation for the reduction (a) and oxidation (b) of water, respectively, under hydrogen and oxygen 1 atm pressure. Lines 1 and 2 delimit the regions for the equilibria between H 2 C0 3 , HC0 3 - , and C0 3 2 - in aqueous solution free from
-............
". ..----.b ~ CO2 ................................
~
H2C0 3
~
HCO;
(ij
E Q) (5 a. 0
".
~-·-··,-· ...C
CO~-
3
........................
CH 30H CH 4
-1 -2
0
2
4
6
8
10
12
14
16
pH
Figure 19.2 Potential-pH equilibrium diagram for the system C(graphite)-water at 25°C and for log(concentration) or log(partial pressure) equal to zero. (Reproduced from Ref. [3] with permission from Elsevier).
19.3 Relevant Characteristics of Carbon Electrode Materials
oxidizing agents. The domain above lines 3, 4, and 5 corresponds to solutions containing 1 M of dissolved carbon in the form of H 2 C0 3 + HC0 3 - + C0 32(corrosion region). The domain below these lines refers to solutions saturated with solid carbon in equilibrium (immunity or stability region). Above line (b) CO 2 is the stable form of carbon. For log c = logPcH4 = logPcoz = 0, carbon in the form of graphite is thermodynamically stable only over a limited domain. Thermodynamic data have also been calculated for carbon-oxygen reactions in fused salts [7, 8]. The oxidation of solid carbon principally yields carbon dioxide at low temperature and carbon monoxide at high temperature. In this case, at constant temperature, the CO/C0 2 concentration ratio at solid carbon depends on pressure. The carbon-oxygen electrode is used as reference to investigate cryolite-alumina melts at c. 1000°C [9] and molten slags at higher temperatures. Thermodynamic data for other systems involving carbon and carboncontaining compounds are given in the original publications [3, 6, 10].
19.3
RELEVANT CHARACTERISTICS OF CARBON ELECTRODE
MATERIALS
19.3.1 Types of Carbons Used in Electrochemistry Carbon has been widely used since the times of Humphrey Davy (17781829), who used charcoal electrodes in some of his experimental work [11]. Carbon electrodes are extensively employed in a large number of electrochemical processes [12, 13], including electrochemical energy storage and energy conversion devices, halogen production, electrometallurgical processes in melts and aqueous solutions, water preparation and water decontamination systems, preparation of organic compounds by chemically modified electrodes, as well as inorganic electrosynthesis to generate peroxide, ozone, fluoride, chloro-alkali, and metals from fused salts [14, 15]. Carbon and graphite are often used as supports for electrocatalysts, but they also have an electrocatalytic function in electrode reactions such as oxygen reduction in alkaline electrolytes, chlorine alkali industry, and SOCl2 reduction in lithium-thionyl chloride batteries. Carbon electrodes are also employed in electroanalytical applications due to the very low residual current over a wide range ofpotentials that makes it possible to study electrochemical reactions even at the level of trace concentration. Among the different types of such electrodes, wax-impregnated graphite rods, carbon powder bound with an inert viscous liquid (carbon paste), glassy carbon, pyrolytic graphite and carbon fibers, and, more recently, nanotubes and fullerenes can be mentioned. Carbon fibers have radial, random, or anion distributions that lead to a different distribution of step and step-step interactions.
486
Chapter 19 Electrochemical Behavior of Carbon Materials
19.3.2 Structural Aspects Carbon in the form of graphite behaves as a good metal. In the form of diamond it constitutes a wide-gap super hard semiconductor; with the intercalation of appropriate guest species it turns into a superconductor [16]; as a flexible polymer it reacts with hydrogen and other species. Carbon-based electrode materials show the entire range of dimensionalities (D) from fullerenes (OD quantum dots), to carbon nanotubes (lD quantum wires), to graphite (2D layered anisotropic material), and to diamond (3D wide gap semiconductor). Graphite represents the ground state for a system containing a large number of carbon atoms. Each small graphite sheet has a large energy per carbon atom at edge sites. In contrast, a small number of carbon atoms form closed shell configurations as in fullerenes and carbon nanotubes [1 7, 18]. The tunneling conductance between neighbor carbon nanotubes can be uniquely specified in terms of their individual chiral vectors and the pentagon and heptagon that must be introduced in the junction region. The conductance between two metallic nanotubes is found to be ballistic with some reflection effects occurring in the junction region. A metal semiconductor nanotube junction shows tunneling across the junction [19]. Results from scanning tunneling microscopy (STM) measurements indicate that one metallic nanotube 8.7 nm in diameter exhibits an ohmic behavior, whereas semiconducting tubules 4.0 and 1.7 nm in diameter show plateaus at zero current passing through null voltage. The slope of the current vs voltage plot provides a measure of the density of states. The current peak heights in these plots depend on the square root of the energy gap-dependent singularities in the lD density of states. Semiconducting tubules show a linear dependence of their energy gap on the reciprocal tubule diameter [20]. The electronic and phonon dispersion relationships for pristine graphite have constituted the basis of models for other less well-ordered forms of graphite such as disordered graphite, graphite intercalation compounds, and ion-implanted graphite [16, 21]. The electrical resistance of carbon increases with oxygen chemisorption at the surface. Powdered carbon reactions with oxygen at SOO-700°C result in a 4% oxygen content and in a 100-fold increase in the electrical resistance [22].
19.3.3 Surface Free Radical States Electron paramagnetic resonance (EPR) is of considerable value for identifying paramagnetic surface groups and clarifying their role in electrochemical reactions. The variety of EPR characteristics of carbon and graphite reflects the diversity of structural and electronic properties of these materials that depend on crystalline size and perfection, impurities, preferred orientation, electrical resistivity, physical adsorption of gases, preparation procedure, and measuring techniques [23]. The surface of carbons is characterized by their capability for oxygen chemisorption at low temperatures. Well-defined crystalline graphite exhibit well-ordered stacks of carbon layers that are fairly unreactive toward
19.3 Relevant Characteristics of Carbon Electrode Materials
oxygen chemisorption, in contrast to more disordered structures such as carbon blacks yielding carbon-oxygen surface complexes. Free radical states have an important role in the surface chemistry of carbons. They are formed as a result of thermal splitting of the C-H bonds to produce carbon rings. Unpaired electrons stabilize by occupying a molecular orbital in the 7T-bond system. The ratio between the electron density in the 7T-bond system and conduction electrons depends on temperature and on the treatment of the material. Polyconjugated carbon structures that provide 7T-electrons usually involve three kinds of free (7-radicals: single radicals, side radicals, and (7-radicals without participation in the conjugated system. The electron capture by the broken (7-bonds is more favorable than that by the 7T-bond as the corresponding energy difference is about 403]/mol. This fact leads to a variety of primary oxygencontaining surface states yielding the appearance of carbonylic, carboxylic, hydroxylic, and quinone groups at the edges of carbon layers. Hydrogencontaining groups are also formed, as demonstrated by surface analysis. These surface states affect the chemical and electrochemical properties ofcarbon surface [24]. The amount of carboxylic and phenolic groups can be determined from the amount ofnitrogen produced by their reaction with diazomethanes (19.15)
(19.16) The distinction between these groups can be made by reaction of the carboxylic group with HCI [25] R-COOCH 3 + HCL --* R-COCI + CH3 0H
(19.17)
Quinone groups can be quantitatively determined from the amount of hydroquinones that is produced by reaction with NaBH 4 [25]. Lactones exhibit IR bands at 1760 cm- 1 because of the CO group of a lactone, and at 1600 cm- 1 because of the CO group hydrogen bonded to a phenolic OH. The band at 1600 cm -1 disappears upon formation of the sodium salt. The surface of carbons can be modified from hydrophobic to hydrophilic by means of oxidation processes. Consequently, carbons can exhibit selective adsorption properties depending on their oxygen content. For instance, commercial carbon blacks with a significant oxygen content selectively adsorb methanol from a methanol/benzene mixture, whereas one with much lower oxygen content exhibits selectivity for benzene.
19-3-4 Double-layer Properties The capacitance-potential curves of the basal plane of highly ordered pyrolytic graphite (HOPG) (Fig. 19.3) show an anomalous low capacitance
488
Chapter 19 Electrochemical Behavior of Carbon Materials
4.0 .----.....----....-----......-.....
N 3.0 E
~
.6 (,)
2.0
0.5
o
-0.5
E(V)
Figure 19.3 Capacitance-potential curves for HOPG in NaF solutions of pH of about 6 at 25°C; a.c. measurements at 20 Hz. (Reproduced from Re£ [26] with permission from Elsevier.)
value that is in the range 1.9-3.0 J.LF/cm2 , depending on the electrolyte solutions concentration [26]. It exhibits a negligible frequency dependence, both in acid and in base, and is nonsensitive to the presence of iodide in the solution. These facts indicate that the surface is free of functional groups to interact with the ions. The low capacitance value for the basal plane is related to a space charge caused by the semimetal characteristics of HOPG. The capacitance-potential curve of graphite is essentially parabolic rather than hyperbolic, probably because of imperfections on the exposed basal plane giving rise to sites with degenerated surface electronic states. Thus, the capacitance calculated from slow scan voltammetry is about 29 J.LF/cm2 at -0.2 V (vs normal hydrogen electrode (NHE)), a value considerably higher than that obtained from alternating current impedance measurements. This suggests that the much larger capacitance represents a portion of the surface with a micro-orientation that exposes other than the basal plane, or it might correspond to the possible existence of microfissures or microvoids. Exposed edge orientations have a much higher capacitance of about 60 J.LF/ cm2 that adds a large resistive component in series arising from the electrolyte resistance. Similar conclusions have been derived from glassy carbon [27]. Conversely, the voltammogram of HOPG in 0.5 M aqueous H 2 S0 4 and 1 M aqueous NaOH at 25°C is relatively featureless
19.3 Relevant Characteristics of Carbon Electrode Materials
at least in the range 0-0.75 V (vs NHE), in agreement with the features of the capacitance/potential curves [28]. The potential distribution across the carbon-electrolyte solution interface in general will be changed by the surface functional groups. Correspondingly, the oxygen-containing groups may influence the potential of zero charge and the potential at the outer Helmholtz plane (OHP) of the electrical double layer [1]. Thus, even for the redox species that are not specifically adsorbed, their concentration at the OHP would be changed and this would also affect the kinetics of the reaction. Potentials of zero charge of various types of carbons in aqueous solutions are in the range 0.0-0.32 V (vs NHE) [6] Black carbons form a homogeneous material series with graphitized black as the reference. For this series the chemical response ranges from Lewis baselike to Brnsted acid-like, while the work function varies appreciably through a minimum across a seven order of magnitude variation in the aqueous solution pH. The decreasing portion shows the lessening influence of the Lewis basic-like carbon basal plane electronic structure as acidic localized oxide functionalities are added to the carbon surface. The subsequent increase in the work function for pH < 6 is attributed to the accumulation of an outwardly pointing surface dipole layer with electric dipoles of2.6 D associated with the stronger (carboxyl) acidic functionalities. The work function measurement has been made using the Kelvin-Zisman reciprocal capacitor technique that consists of determining the contact potential difference between the carbon black and a gold reference surface. Values of the work function are in the range 0.19-0.30eV [29]. Capacitance measurements of carbon electrodes have also been made in molten halides, particularly chlorides [30-32], molten nitrates [33, 34], and in cryolite-alumina melts (graphite and glassy carbons). In cryolite-alumina melts, the double-layer capacitance of the basal plane of graphite, in the range 0.7-1.0 V (vs aluminum reference electrode) is about 20 f.1F/ cm2 at 0.9 V, i.e., in a potential range where no appreciable flow of current has been observed. Data indicate that the capacitance is influenced by adsorbed species from the melt, possibly yielding intercalation compounds, and uncertainty in the true area of the electrode [34].
19-3-5 Roughness Factor Carbon surfaces, except the HOPG basal plane, have some degree of roughness. The roughness factor (a) can be defined as (19.18) On the assumption that the surface roughness is on a distance scale which is large compared to the analyte molecules, Am is the microscopic area that is relevant for adsorption or kinetic measurements. A g is the geometric area determined either visually or by chronoamperometry on a scale where J15:i is much greater than any surface roughness. D; is the diffusion coefficient of the reactant
49°
Chapter 19 Electrochemical Behavior of Carbon Materials
i in the ionic conductor and t the electrolysis time. The value of (J" > 1 refers to the entire microscopic area disregarding the amount or distribution ofedge planes. For the edge plane area (A edge )
(19.19)
Ie represents the fraction of edge planes on the surface, and depends strongly on the nature and preparation of the carbon surface. The roughness of carbons is sensitive to the applied potential routine, as seen by sequential nanoscopic images of HOPG surfaces in aqueous solutions subjected to potential cycling of different duration [35] (Fig. 19.4). A stabilized carbon electrode topography merges after a prolonged potential cycling. These topographic changes can be described as time effects that depend on the type of carbon and ionic conductor, and the characteristics of the current or potential perturbation routines [20].
19-3-6 Fractality The problem of transfer across a fractal surface has been considered in the electrochemical behavior of rough and porous carbon electrodes [36]. The fractal dimension can be determined from nitrogen gas adsorption data, from transmission electron microscopy (TEM), and nanoscopy image analysis. Fractal electrodes exhibit a constant phase element (CPE) behavior in electrochemical impedance spectroscopy (EIS) [37]. The relationship between the CPE behavior of rough, irregular electrodes and fractality depends on the scale of irregularities, i.e., whether it is on the micrometer or centimeter scale. In real situations, however, both microscopic and macroscopic geometric effects probably occur simultaneously. For imprinted mesoporous carbons, the overall fractal dimension, determined from gas adsorption data, indicate that these materials are composed of two groups of pores. The surface fractal dimension of the carbonization-induced pores surface and that of the silica-imprinted pores surface has been obtained from TEM image analysis [38].
19-3-7 Intercalation of Ions in Graphite Intercalation constitutes an important case of inclusion phenomena in which the host lattice is characterized by a lamellar structure [39]. Graphite yields both anion and cation intercalation compounds and charge transfer processes are the driving forces for their formation. Due to the action of an oxidizing agent electrons are drawn from the graphite lattice and anions beside neutral species are intercalated. These processes can be driven in a direct reversible electrochemical way, as has been demonstrated for carbon in concentrated sulfuric acid [39]. For a graphite electrode in concentrated acid solution, the formation of intercalation compounds occurs when the threshold potential for the intercalation
19.3 Relevant Characteristics of Carbon Electrode Materials
49 1
Figure 19.4 Constant-current scanning tunneling microscopy images (600 x 600 nm2 ) of HOPG after anodic oxidation in 0.1 M H 2 S0 2 at 0.05 V vs Ag/Ag+ electrode. (a) HOPG surface before electro-oxidation (blank); (b) HOPG surface after 20 potential cycles; (c) HOPG surface after further electro-oxidation cycles. (Reprinted with permission from Ref. [35]. © 1988 American Chemical Society.)
Chapter 19 Electrochemical Behavior of Carbon Materials
49 2
process is exceeded (intercalation overpotential). The first step of this process is the oxidation of graphite to form a macroradical cation (19.20) the electron being removed from the highest filled level of graphite. This process resembles the anodic oxidation oforganic hydrocarbons such as perylene yielding a radical cation. In both cases, the anion acts as the counterion required to balance the positive charge. The second step is the transfer of anions across the electrochemical interface (19.21) and correspondingly, the graphite lattice has to be expanded. For weakly solvated cations of low melting points, this process is highly reversible, as has been concluded from voltammetry and impedance measurements [39]. For carbon in concentrated sulfuric acid, the overall reaction can be represented as follows: (19.22) From reaction (19.22) intercalation compounds (i) with x = 24 have been obtained. Alkaline metals, particularly rubidium, potassium, and cesium, intercalate graphite layers yielding compounds of the form CsMe when a layer of alkaline metal atoms is formed between each pair of carbon planes [6].
19.4
CHEMICALLY MODIFIED ELECTRODES AND
SUPRAMOLECULAR CONFIGURATIONS
The electrochemical and electrocatalytic properties of carbon electrodes can be modified changing their surface composition by anchoring foreign compounds. This can be accomplished by adsorption, by chemical reaction with a surface group, by specific chemical binding, and by adsorption immobilization on a sublayer of a polymer material [40]. Chemically modified electrodes constitute a part of supramolecular chemistry. Typical examples of the adsorption procedure are the irreversible attachement of metal-N4 complexes on HOPG and the adsorption of aromatic molecules for anchoring complex species. Ion-N4 complexes are adsorbed in a planar orientation on the HOPG surface and cobalt and iron tetrasulfate phtalocyanines are arranged sideways relative to the surface [41].
19.4 Chemically Modified Electrodes and Supramolecular Configurations
493
Complexes like [Ru(NH 3 )sL]2+ with a large aromatic ligand such as 4aminomethylpyridine or N-(4-picolinic)benzamide; [(RubipY)2L]2+·2(PF 6 -); 1,5 diihydroxyanthraquinone, can be adsorbed on glassy carbon by evaporation from a nonaqueous solution [42]. For chemical attachment, the carbon surface is first activated by oxidation at 160°C in air or by oxygen plasma. Then, activated COOH carbon groups react with thionyl chloride yielding -COCl groups at the carbon surface. Subsequently, the active group (R) is attached via a reaction with amines leading to -CONHR. Thus, different functional groups (R) can be attached. Covalent attachment of active molecules to graphite surfaces can be made via OH groups using cyanuric chloride, 2,4-dinitrophenylhydrazine or chlorosilanes as intermediate reagents [43]. The adsorption of polymers, poly(vinyl pyridine) or poly(acrylonitrile) either to coordinate metal atoms or to adsorb biopolymers has been used to prepare chemically modified electrodes for immobilization of enzymes either by physical or by chemical adsorption (carrier binding), cross-linking, and entrapping at lattice sites or in microcapsules [43]. A wide application of these types of electrodes has been made for electrochemical reactions ofbiological interest [44]. Chemically modified electrodes resulting from the attachment of quinones, phenantroline, dipyridyl complexes, and N 4 complexes, from the development of polymer-coated carbon materials, and from electrodes modified by enzymes have been specifically designed for the electrocatalytic reduction of molecular oxygen (OERR). Carbon materials with immobilized hydroquinone have also been utilized to accelerate the electrochemical oxidation of molecular hydrogen. Modified carbon and graphite electrodes have been found adequate for producing a mixture ofoptically active isomers and stereoselective addition reactions such as the chlorination of anisole at an a-cyclodextrine-modified graphite electrode [45]. The kinetics of the electrochemical reactions at arrangements of chemically modified electrodes has been interpreted by a charge and mass transfer electrochemical mechanism. Charge transfer can be, in general, described by an electron jump and a molecular diffusion step. At electrodes modified by complexes, the rate of electron tunneling (W(r)) can be described by the equation
W(r) = V' exp (-r/ A')
(19.23)
r being the distance covered by the electron, V'is a constant, and A' depends on the geometry of the potential energy barrier. Accordingly, the transfer efficiency should depend on the distance from the active center to the electrode plane. For r < rerit' W(r) should be greater than the rate of the reaction at the active center. For r > rerit' the reverse situation occurs. The influence of ron the rate of the OERR has been studied on laccasemodified carbon electrodes [46]. In this case, rwas varied by a monolayer of adsorbed lipid that had either planar (cholesterol) or vertical (lecithin) orientation on the electrode surface. In this case, a sharp decrease in the rate of OERR was found within a narrow range of r, which is determined by rerit ~ 2 nm.
494
Chapter 19 Electrochemical Behavior of Carbon Materials
The diffusion step becomes important for polymer-modified electrodes. Thus, the apparent diffusion coefficient depends on the concentration of redox groups because the acceleration of the electron exchange decreases with the ion distance. These conclusions were drawn from a series of polyvalent ions anchored electrochemically to poly(4-vinyl pyridine) on graphite [47]. Electronic conductivity is favored by electron transfer through the polymer delocalized band structure, via redox conductivity by site-site hopping. Redox conductivity occurs at electron energies centered around the formal equilibrium potential for the redox polymers.
19.5
ELECTROCHEMICAL KINETICS ON CARBON
ELECTRODES IN AQUEOUS SOLUTIONS
19.5.1 Direct Electrode Processes Although carbon electrodes are frequently used for electroanalytical studies of oxidizable compounds, many of them exhibit heterogeneous charge transfer rates that are very low at carbon electrodes, as concluded from their corresponding ill-defined voltammograms [48]. Thus, the surface properties of carbon electrodes can have remarkable effects on the voltammetric response of these direct electrode reactions. One typical example of this behavior is the voltammogram of the ferro/ferricyanide couple (test reaction) that at carbon electrodes is less reversible than at noble metal electrodes. The kinetics of the test reaction in 1 M aqueous KCI was used as the reference to compare its electrochemical behavior on different carbon electrodes [20]. This electrochemical reaction occurs via an outer sphere mechanism and its rate depends on the electrolyte composition and can be increased by appropriate treatment of carbon electrodes, for instance, by application of a high current potential routine to electrodes of carbon fibers. Similar results have been obtained with glassy carbon surfaces that had been pretreated at SOO°C under reduced pressure. An alternative activation method is based on careful electrode surface polishing [6]. The kinetics of the test redox reaction on the cleavage HOPG surface is almost under pure diffusion control, whereas on the edge surface, where ionspecific adsorption is favored, it is under combined kinetic and diffusion control. Accordingly, surface heterogeneity is a new variable in the kinetics of electron transfer processes at carbon surfaces, as the surface energy of sites at each domain (plane, edges, kinks, etc.) is different. Let us assume the existence of two surface domains (1 = basal and 2 = edge) for graphite with specific rate constants for the test reaction (k~ and k~) and consider that the 1-1 and 2-2 domain distances are R 11 and R 22 , respectively. The value of these distances relative to ffi may have a significant effect on the voltammetry that depends on whether R 11 , R 22 «ffi, R ll < ffi < R 22 , and R ll , R 22 »ffi. The value
19.5 Electrochemical Kinetics on Carbon Electrodes in Aqueous Solutions
495
°
of kg = 0.10 cmls is about 25 times greater than k 1 = 0.004 cmls, in agreement with the difference in reactivity between the basal plane of graphite and the edge sites. The remarkable anisotropy of kO coincides with that of C, E pzc and the work function (Section 19.3.4). Carbon composites have been developed as alternative materials for carbon paste electrodes because of the limited utility of the latter in most organic solvents. These composites include polyethylenelcarbon black [49], Kel-F/graphite [50], carbon black immobilized in cross-linked polyethylene [51], and epoxy/graphite [52]. A collection of kO for these materials is available [20]. Thus, for platinum kO = 0.24 cm/s, for pyrolytic graphites 0.002 < kO < 0.007 cm/s, and for graphite carbons 0.005 < kO < 0.14 cm/s. Simple redox solutes, ferrocene, N, N, N, N -tetramethyl-1 ,4-phenylenediamine, decamethylferrocene, bis(i-propylcyclopentadienyl) iron(II), [Ru(phen)3] (CI0 4)2' [Fe(bpY)3] (CI0 4)2' [Co(bpY)3] (CI0 4)2' and iodine have been studied at electrodes modified with polymeric fullerene films. Fullerene-modified electrodes were prepared by electropolymerization of C 60 initiated by traces of dioxygen or by simultaneous electroreduction of fullerene and Pd(II) acetate trimer. For the former films, the electrochemical activity decreases upon potential cycling. The electrochemical activity of the film is stabilized by the redox solute added to the electropolimerization stage due to the catalytic oxidation of the fullerene film by the oxidized form of the redox system. Similarly, positively charged species can also be incorporated into the structure of the film. The reversible behavior of redox solutes decreases with the increase in the thickness of the Pd/C 60 film. This film also incorporates ferricinium ion, N, N, N, N -tetramethyl-1 ,4-phenylenediamine cation, decamethylferricinium ion, and to a smaller degree [Co(bpY)3]n+ [53]. Microcrystals of fullerene-C 6o on glassy carbon mediate the oxidation of cysteine in the presence ofaqueous potassium-containing electrolytes. The potential for the oxidation of cysteine is lowered by approximately 100 illV and current is enhanced significantly as compared to bare glassy carbon electrodes. Additional mediation occurs when the potential range of C 60 1C 60 n- redox couples are covered. The electrochemical response is sensitive to pH, temperature, and C 60 dosage. Excellent analytical and/or recovery data are obtained with vitamin pill (alcovite), cassamino acid (hydrolyzed casein), and for a range of beverages [54].
19.5.2 Oxygen Electroreduction on Carbon Electrodes The reversible potential of the water decomposition reaction is 1.23 V at 25°C. The overpotential for OERR in aqueous alkaline solutions (19.24) is 0.3-0.4 V at 60-80°C, and in acid solutions 02 + 4H+ + 4e- -+ 40Hit is 0.4-0.5 V at c. 190°C
(19.25)
Chapter 19 Electrochemical Behavior of Carbon Materials
The oxygen electrode polarization is a measure of the degree of irreversibility of the electrochemical reaction. To find an effective electrocatalyst for reactions (19.24) and (19.25) is of a great interest because of their technical relevance in water electrolysis, fuel cells, metal corrosion in aqueous environments, biological processes, etc. The OERR is usually considered to proceed via two reaction pathways, namely, the peroxide and the direct four-electron pathways. The scheme of the peroxide pathway is H02(ad) + OH-
H2 0 + O2
~I ~
(19.26)
where (ad) stands for peroxide adsorbed species on carbon. The peroxide species are either electroreduced further to OHHOH + H02-
~ ~
3HO-
(19.27)
or catalytically decomposed to OH- and O 2 2H0 2
(19.28) 2H02" (ads)
The overall reaction is the four-electron electroreduction of molecular oxygen. The oxygen resulting from reaction (19.28) is recycled via reaction (19.26). The direct four-electron pathway involves no hydrogen peroxide formation in the solution. This fact, however, does not preclude the participation of an adsorbed peroxide intermediate in the course of the reaction. The distinction between both reaction pathways is usually investigated by the rotating ring-disc electrode technique [55]. From the rotation speed and potential dependence of the disc electrode to ring electrode current ratio, it is possible to determine the relative contribution of each reaction pathway to the overall reaction [56].
19.5 Electrochemical Kinetics on Carbon Electrodes in Aqueous Solutions
19.5.2.1
497
OERR kinetics in alkaline solutions
The kinetics of the OERR on carbon [27] and graphite [27] in alkaline solution has been explained in terms of the dominant contribution of the peroxide reaction pathway. On the other hand, the direct four-electron pathway predominates on graphite electrodes modified by adsorbed tetrasulfonated phtalocyanine [57] and attached face-to-face di-cobalt-porphyrin complexes [58]. In principle, when both pathways operate simultaneously on a given surface, the kinetics is referred to as involving a parallel mechanism [59]. In alkaline solutions porous carbon electrodes are effective catalysts for the OERR. In this case, the exchange current density of reaction (19.26) for both carbon and graphite in 1 M KOH + 10-3 M peroxide concentration [59] is in the range 10- 4 < Jo < 10-3 AIcm2 (true area). For porous carbon electrodes (10 4 -10 5 cm2 true area per cm2 superficial area) large values ofJo indicate a small activation polarization for the OERR. The predominant process occurs then via the peroxide pathway. In general, the presence of impurities determines the extent of the rate of desorption of adsorbed peroxide, although the catalytic peroxide elimination effect decreases during the electrode operation. For porous structured carbons this effect can be due to the buildup of a substantial amount of peroxide in the solution within pores. Furthermore, the high peroxide concentration contributes to increasing the O 2- radical ion and OH radical concentration within the pores via the following equilibrium: (19.29) the equilibrium constant of reaction (19.29) being K ~ 10-7 .5 at 25°C [60]. Radicals such as O 2 and OH, which are produced as intermediates in the homogeneous peroxide decomposition, may favor the attack of carbon via oxidation. This fact is accompanied by a change in hydrophobicity and porous clogging by gel formation, particularly because of sodium peroxide. Suppression of H0 2- concentration in porous carbon electrodes is usually accomplished by the dispersion in carbon of specific catalysts such as silver, Mn0 2, and Ni-Co spinels. For very active catalysts such as platinum supported on carbon, the direct four-electron and the peroxide-producing reaction occur in parallel, the first on the catalyst, and the latter on carbon surface domains. Accordingly, it is possible to diminish the peroxide activity to the equilibrium value of reaction (19.29). In this case, the electrode potential would approach the equilibrium value for the overall four-electron electroreduction reaction (19.24). The stationary cathodic current-potential polarization curve of the OERR on pyrolytic graphite exhibits the Tafel slope -0.120 V per decade- 1 at 25°C, and the stoichiometric number is 2 for the O 2 to OH 2- electroreduction reaction. For glassy carbon, the Tafel slope is -0.060 V per decade- 1 , and the corresponding stoichiometric number is 1, as expected for reaction (19.26) [28]. The OERR is first order in 02 and zero order in OH- for both carbons.
Chapter 19 Electrochemical Behavior of Carbon Materials
For graphite the following mechanism for the OERR has been proposed [27]: O 2 -+ 02(ads) O 2(ads) + e- -+ O~ 20~(ads) +HOH -+ O 2 +HO~ +OH-
(19.30a) (19.30b) (19.30c)
with step (19 .30b) being rate determining in the Tafel range of the polarization curve. Reaction (19.30c) is a complex process involving several steps. For glassy carbon, the mechanism of the OERR [28] starts with reaction (19.30a) followed by 02(ad)+e- -+ [02(ads)]-
(19.31a)
02(ads)- -+ {02(ads)}-
(19.31b)
{02(ads)}- +HOH-+ H0 2(ads)+OH-
(19.31c)
H0 2(ads) + e- -+ HO~(ads)
(19.31d)
HO~(ads) -+ HO~
(19.31e)
step (19.31c) being rate determining. [02(ads)] and {02(ads)} indicate different adsorbate structures. For Teflon-bonded gas-fed electrodes prepared from carbons that have little peroxide-decomposing activity, the OERR at the highest current densities appears to be limited by converging characteristics related to carbon itself, its electrocatalytic activity for oxygen reduction to peroxide and peroxide decomposition, the gas mass transport, and the electronic conductivity. To advance in the understanding of the OERR mechanism on carbon and graphite, more information at the molecular level of surface functional groups at these cathodes in air is still required. 19.5.2.2 OERR in acid solutions
In acid solution the OERR proceeds mainly via the formation of H 20 2 on porous carbon electrodes. This is also supported by experiments with 18 0 isotope that showed a lack of 0-0 bond break during the OERR. In acid media, the OERR appear to be independent of pH and the Tafel slope is close to -0.120 V per decade 1 , the transfer of a first electron to an adsorbed oxygen molecule being the rds. (19.32) The reaction is first order with respect to molecular oxygen. The efficiency of the OERR is increased considerably when mesometal and nanoparticles (Pd, Au) on carbon surfaces are used as electrocatalysts [61]. This electrocatalytic enhancement is related to the geometry of these metal islands
19.5 Electrochemical Kinetics on Carbon Electrodes in Aqueous Solutions
499
and it appears that the most active domains are located at edges of islands in contact with the HOPG surface [62]. This is consistent with the fact that gold nanoparticles electrochemically formed on graphite are preferentially deposited on the upper plane of step edges due to the nonuniform electron density that results from relaxation of the graphite lattice near steps [63].
19.5.3 Oxygen Reduction on Macrocyclic Transition Metal Complexes on Graphite and Carbon Surfaces In contrast to the interaction of O 2 with graphite and carbon surfaces, the electrodes modified by transition metal complexes provide the possibility of extending the type of interactions derived from inorganic chemistry to the electrochemical system. A typical example is the face-to-face anchorage of porphyrins as catalysts on carbon electrodes for the OERR [58, 60, 64]. For cobalt porphyrin on graphite, when the Co-Co distance is about 0.4 nm, which makes the formation of an 0-0 bridge between Co centers possible, the presence of the Co porphyrin catalyzes the four-electron reaction in acid solutions, whereas for smaller Co-Co distances, the peroxide pathway is catalyzed. These behaviors have been related to the cis and trans surface complex configurations that assist the four-electron reaction and the peroxide pathway, respectively. Similar electrocatalysis for the OERR has been found in alkaline solutions when the macrocycles are adsorbed on graphite [28]. The thin layer of transition metal macrocycles attached to carbon generally lack long-term stability in concentrated acid and alkaline solutions. This drawback can be overcome by thermal treatment at 450-900°C for cobalt tetramethoxy phenyl porphyrin (Co-TMPP) [65]. Under these conditions, the Co-TMPP is substantially degraded to cobaltous oxide. Pyrolyzed layers involve high-area carbonaceous materials with a significant surface nitrogen and the transition metals as small oxide and metallic particles dispersed on the high-area substrate. These layers catalyze peroxide elimination in alkaline solutions. The catalytic current for the OERR in aqueous solutions at glassy carbon electrodes modified by the physical adsorption of 1,2-dihydroxyanthraquinone is significantly increased under insonization because of the increase in mass transport [66].
19.5.4 Oxygen, Hydrogen, and Chlorine Electrode Reactions Hydrogen, oxygen, and chlorine overpotential measurements on the basal and edge planes of stress-annealed graphite are complicated by intercalation and oxidative attack of the surface. Both hydrogen and oxygen overpotential are quite high on most graphite and carbon surfaces, probably in part because of the existence of functional groups. The interaction of adsorbed groups, both directly or through electronic substrate effects, would produce broad voltammetric peaks that reflect in large Temkin terms in the adsorption isotherm. This fact makes the voltammogram interpretation difficult.
Chapter 19 Electrochemical Behavior of Carbon Materials
500
19.5.4.1 Hydrogen evolution on carbon and graphite Carbons exhibit a low electrocatalytic activity for the hydrogen electrode reaction (HER). Structural characteristics have significant electrocatalytic effects on the HER as Jo changes from 2 X 10-9 to 2.5 X 10- 8 A/cm- 2 on going from the basal plane to the side face of pyrolytic graphite. On glassy carbon, the HER overpotential decreases as the pretreatment temperature is increased. This thermal treatment leads to structural and chemical transformations from carbonization, precrystallization, and to graphitization. Kinetic parameters for the HER on different carbon and graphite electrodes show that depending on the type of electrode and acid solution Jo varies between 2 x 10- 7 and 2 x 10- 13 A/cm2 and the cathodic Tafel slope is usually close to -0.120 V per decade, although some unexpected higher values have also been recorded. The reaction order with respect to the hydrogen ion concentration is 1, but unexpected values of 0 and 2 have also been reported [6]. The rds is usually the initial discharge step and the surface coverage ofhydrogen atoms is low. Electrochemical reductions of fullerene films in the presence ofBrnsted acids yield hydrogenated fullerenes H n C 60 , where n depends on the acid, its concentration, and on the electrode potential. Hydrogenated fullerene films behave as semiconductors with increased photoefficiency [67].
19.5.4.2 Oxygen evolution on carbon and graphite The rate of the oxygen evolution reaction (OER) on pyrolytic graphite is higher than that for glassy carbon. For both the carbon electrodes, the temperature pretreatment has no influence on the current measured at constant potential. Carbon dioxide is the main reaction product for E < 1.1 V (vs reversible hydrogen electrode (RHE)) on pyrolytic graphite. For a pH between 1 and 9, the Tafel slope changes from 0.150 to 0.240 V per decade, depending on the solution composition and electrode preparation. The anodization of both glassy carbon and HOPG in aprotic solutions (DMSO, ACN) is characterized by a reversible one-electron O 2 to 02 - reaction. Kinetic data of the oxygen electrode on carbon materials have been compiled in Ref [68].
19.5.4.3 Chlorine electrode In aqueous solutions, the equilibrium potential for the reaction
(19.33) is 1.359 V vs NHE at 25°C [3]. This figure is approached for smooth pyrolytic graphite in aqueous NaCI (a = 1) under chlorine saturation (PC1 2 = 1) to attain 1.320 V [6]. Therefore, carbons are useful for applications in the chlorine evolution reaction as both, the carbon oxidation reaction and the OER exhibit larger overpotentials.
19.6 Organic Electrochemistry at Carbon Electrodes
501
The kinetics and mechanism of the chlorine evolution reaction in aqueous solutions have been studied on smooth, porous, and impregnated graphite [68, 69]. The Tafel slope depends also on the nature and history of carbons. For HOPG and glassy carbon, the anodic Tafel slope is about 0.060 and 0.120 V per decade at 25°C, respectively, whereas for a graphite electrode consisting of a section parallel to the c-axis, three regions in the polarization curve with anodic Tafel slopes from 0.060 to 0.160V per decade have been observed. For reaction (19.33), current flow for the porous electrodes is under ohmic regime. Specifically adsorbed anions hinder chlorine evolution, in contrast to cations such as Fe3+ that probably produce a change in the potential distribution at the electrical double layer. The residual gas evolution at graphite, after switching the anodic current 0[£ decays by desorption via self-discharge. graphite(Cl)ad + e- ~ ClCl- + graphite(Cl)ad ~ C12 + e-
(19.34) (19.35)
The rate of diffusion of atomic chlorine is determined by an equilibrium between diffusion ofadsorbed chlorine from graphite outward (Eqn. (19.34)) and the formation of molecular chlorine (Eqn. (19.35)). As the surface concentration of chlorine diminishes, it is replenished by diffusion, a process that gradually becomes rate determining. The kinetics of molecular chlorine evolution follows a first-order law with respect to chlorine, and zero order with respect to chloride ion concentration. The chlorine impregnation of carbon electrodes results in lamellar compounds such as CsCl [6]. The cathodic Tafel slope corresponding to reaction (19.33) in the reverse direction is close to -0.120 V per decade at 25°C. For both cathodic and anodic reactions, the interfacial capacity results in 30-35 f.1F cm2 . This figure is consistent with a low surface coverage by chlorine atoms. Kinetic parameters for the chlorine evolution reaction on carbon electrodes are assembled in Ref [6].
19.6
ORGANIC ELECTROCHEMISTRY AT CARBON
ELECTRODES
Carbon, graphite, vitreous carbon, carbon felt, carbon fibers and cloth, as well as reticulated carbons, are of common use either as anodes or cathodes for the electrochemical synthesis of organic compounds. A typical example is the Kolbe electrosynthesis in which a carboxylate salt is electrochemically discharged at vitreous carbon anodes, in both aqueous and nonaqueous media, yielding a hydrocarbon with high efficiency and carbon dioxide.
Chapter 19 Electrochemical Behavior of Carbon Materials
502
At ordinary graphite electrodes in aqueous solutions, the reaction products are those derived from the formation of carbonium ion intermediates, RCO~
~
R+C0 2 +e-
(19.36)
2R
~
R-R
(19.37)
~
R++e-
(19.38)
R
H2 O
R+ ~ ROH+H+
(19.39)
RCO-
R+
~2 RC0 2R
-H+
R + ~ olefin or rearranged species
(19.40) (19.41)
The difference in the yields of products appears to be a carbon surface area effect that acts as product-determining characteristics. For an extensive description of this matter, see Ref [15].
19.7
REACTIONS ON BIOLOGICAL ACTIVE ELECTRODES
Electrochemical reactions ofa large number ofbiological active compounds such as aminoacids, proteins, catecholamines, alkaloids, purines and their nucleosides, NAD, FAD, FMN, and nucleic acids have been investigated on HOPG, paste electrode, graphite, glassy carbon, carbon fibers, and fullerenes [68]. In this case, the ability of the anchored compound to remain stable by repetitive potential cycling between its different oxidation states is essential to successfully design a supramolecular electrode for this particular type of electrocatalytic reactions. One example of this type of electrodes is a gold electrode covered by a selfassembled monolayer of gluthation and covalently bound fullerene [70], that has been proposed for the consecutive electro-oxidation of nicotinamide adenine dinucleotide (NADH) to NAD+. A first high power density of about 100 J.1 W 1cm2 miniature biofuel cell uses supramolecular modified carbon fiber electrodes operating in aqueous solution at pH 5 and room temperature. The electrocatalytic film at the anode catalyses the electro-oxidation of glucose to gluconolactone, and at the cathode catalyses the electroreduction of oxygen to water. The supramolecular ensemble involves OS2+ IOs3+ centers and enzymes that are immobilized in the electron-conducting redox-polymer films. The film of the anode consists ofa cross-linked electrostatic adduct of glucose oxidase, and at 0.1 V (vs Agi AgCI) a redox potential electronconducting redox polymer, which electrically connects the glucose oxidase redox centres to one fibre. The film ofthe cathode consists oflaccase and a 0.55 V
19.8 Corrosion Processes
503
(vs Agi AgCI) electron-conducting redox polymer, electrically connecting the laccase redox centres to the second fiber [71]. A miniature biofuel cell operating at 37°C in a glucose-containing aerated physiological buffer consists of two electrocatalyst-coated carbon fibers. Glucose is electro-oxidized to gluconolactone on the anode fibers and dissolved oxygen is electroreduced to water on the cathode fiber. The power output of the cell operating at 0.52 V is 1.9 f.1 W, i.e., a power density of 4.3 f.1 W Imm2 [72]. These advances would increase the probability of achieving technical devices such as sensors, reactors, and energy storage and energy conversion devices, resulting from engineering and design approaches converging to the efficiency found in natural systems.
19.8
CORROSION PROCESSES
The degree of carbon corrosion .depends on the type of carbon, the electrode potential, the temperature, and the carbon pretreatments that affect its surface structure [73]. Corrosion reactions occur at distinct domains of the carbon surface with different rates. The main surface domains are the plane boundaries or defects, outer interplanar areas, and intercalation areas between the planes. The stronger the edge attack, the greater the amorphous domains of carbons. Intercalation between the planes becomes more important with HOPG provided that some edges are exposed to the environment. The corrosion process changes with pH [74]. Nonbasal dislocations play an important role in the oxidation of graphite carbons. In hot 96% phosphoric acid at 135-160°C different carbons exhibit similar corrosion behavior as a function of time [75, 76]. At constant potential, the corrosion current that was initially relatively large decreases rapidly with time. The principal corrosion reaction is (19.42) Different surface oxides are formed as intermediate oxidation products in reaction (19.42). Both the formation of surface oxides and the evolution of carbon dioxide decrease with time. But as the surface coverage by oxide increases, carbon dioxide formation prevails and proceeds via surface oxides at preferred sites. Corrosion rates of carbons appear to be independent of water content and carbon dioxide partial pressure. In acid electrolytes, the Tafel slope for the carbon corrosion reaction appears to be indicative of the degree of disorder on the carbon surface. The larger the Tafel slope, the greater the degree of disorder. The influence of heat treatment on the corrosion rate depends on the structure of the parent carbon, particularly on the lattice parameters. Thus, in hot phosphoric acid at cathodic potentials, as
Chapter 19 Electrochemical Behavior of Carbon Materials
504
used in the phosphoric acid fuel cell technology at 150-200°C, samples of heattreated Vulcan XC-72R after boron doping and heat treatment at 1000-2000°C show an enhanced resistance to corrosion. Changes in Brunauer, Emmit, Teller (BET) surface areas, lattice parameters, and electrochemical behavior converge to show that the addition of boron results in an additional graphitization to that achieved by the heat treatment itself Boron acts as an electron acceptor and can enter the graphite lattice by substituting carbon atoms at trigonal sites that would provide traps for metal clustering. The corrosion of carbon in alkaline solutions is of interest for alkaline batteries. For acetylene black electrodes at 0.45-0.60 V (vs Hg/HgO) in concentrated aqueous KOH, carbon dissolution to carbonate ion, gasification of carbon to carbon monoxide, and oxygen evolution are the main anodic processes, although the potential and temperature dependence of these processes is different. Correspondingly, for E < 0.50 V and T < 50°C, carbon dissolution is the primary process; for 0.50 < E < 0.60 V and T> 50°C, carbon dissolution and oxygen evolution occur at comparable rates; for E > 0.60 V and T > 60°C, oxygen evolution and carbon gasification are the dominant processes. The current efficiency of these processes also depends on whether a catalyst such as cobalt oxide has been added to the carbon electrode, although the major effect is produced on the OER.
19-9
(ARBON ELECTRODES IN MOLTEN SALTS
Carbon electrodes are crucial for a number ofimportant processes in molten salt electrometallurgy. A long list of commonly used metals, such as aluminum, sodium, potassium, calcium, and magnesium are produced by these processes [77, 78].
19.9.1 Cryolite-Al 2 0 3 Melts Carbon anodes are used in the electrolysis of cryolite-alumina mixture to produce aluminum. The overall reaction in the electrochemical cell is (19.43) in which the anode is partially oxidized to CO and CO 2 • The overall anodic reaction is (19.44) However, considering reaction (19.43), reaction (19.44) has been interpreted by a complex reaction pathway that includes the formation of a C(O) surface intermediate as primary process (19.45)
19.9 Carbon Electrodes in Molten Salts
505
followed by a secondary chemical process yielding carbon oxides
C(O) --+ carbon oxides
(19.46)
Another interpretation considers the formation of a nonstoichiometric surface oxide as primary process (19.47) that subsequently decomposes into CO and CO 2 (19.48) The complex mechanism of reaction (19.44) is probably controlled by diffusion and the rate of the heterogeneous chemical reactions. For various carbon electrodes in cryolite melts saturated with alumina at 1010°C, values of 0.0048 < jo < 0.24 AI cm2 have been reported [79]. The overpotential of the anodic reaction increases when the concentration of alumina in the melt decreases, and the wettability of the electrode by the melt decreases because of the accumulation of C(O), CO, and CO 2 species, leading to the dangerous anodic effect. Intercalation compounds such as (CF)n' (C 2 F)n' and C x F(AlF 3 )y are also formed. Compounds having a covalent bond are unique in their low surface energy. The noncovalent intercalation compound results in a conductor better than the original carbon. These findings provided new methods for water proofing the carbon surface and for new materials to be used as cathodes in lithium batteries.
19-9-2 Halides-containing Melts Fluorine is produced by electrolysis of molten salts on carbon anodes including KF-2HF at about 100°C, potassium bifluoride at about 250°C, and fluoride salts at about 1000°C. The decomposition potential ofmolten potassium bifluoride is 1.75 V at 250°C, a value close to that estimated thermodynamically [80]. The kinetics of the anodic process is characterized by a Tafel slope of 0.56 V per decade, jo = 1 X 10-4 A/cm2 [81], and by a complex reaction mechanism involving the formation of fluorine atoms on carbon. During the electrolysis, C-F surface compounds on the carbon anode are formed via side reactions. Intercalation compounds such as (CF)n contribute to the anodic effect in the electrochemical cell, which can be made less harmful by addition of LiF. The kinetics ofthe chlorine electrode in different chloride melts was studied in the range 190-430°C. Different controlled processes involving the participation of chlorine atoms on graphite have been proposed [82, 83]. The evolution and dissolution of chlorine at graphite electrodes was studied in molten lithium chloride. The anodic evolution involves a fast discharge of chloride ions followed by the combination of chlorine atoms that is the rds
50 6
Chapter 19 Electrochemical Behavior of Carbon Materials
of the process. The graphite surface is appreciably covered by chlorine species under Temkin adsorption conditions. The cathodic dissolution of cWorine is limited by diffusion of cWorine in the melt to the electrode surface under high current conditions, whereas at low currents the process of dissociation of chlorine on the surface is followed by the charge transfer process [83].
19.9.3 Oxygen-containing Melts Deposition of carbon from the electrolysis of molten carbonates (Li 2 C0 3 , Na2 C0 3 , and K2 C0 3 ) involves the gradual reduction of the degree of oxidation of carbonate to carbon. At temperatures below 700°C, the formation of carbon is thermodynamically favored compared to that of carbon monoxide [6]. Voltammetry data of graphite electrodes in molten NaN0 3 /KN0 3 at 240-350°C indicate an anodic reaction involving O 2 - ion and NO that proceeds via an oxide group on the graphite surface. The corrosion of graphite was related to the formation of a N0 3 intermediate [84, 85]. For graphite in NaN0 2 /KN0 2 melt at 236°C no appreciable corrosion was observed [86, 87]. The kinetics of the hydrogen electrode reaction on dense porous graphite electrodes in molten KHS0 4 from 245°C to 280°C [88-90] showed that the cathodic and anodic reactions are not strictly conjugated processes. The cathodic reaction was discussed in terms of conventional mechanisms, but the anodic reaction involves the simultaneous oxidation of hydrogen and graphite surface. The reaction exhibits a one-half power dependence on hydrogen pressure. The kinetics of the electro-oxidation of graphite in molten KHS0 4 to volatile compounds (carbon dioxide, carbon monoxide, and traces of sulfur dioxide) was studied in the range 180-320°C. The faradaic yield for carbon dioxide (4 F per mol of carbon dioxide) is about 90 %. The rds is the desorption of oxygen-containing intermediate species [91].
19.10 CARBON ELECTRODE MANUFACTURING TECHNIQUES
Industrial carbon materials are used for molds, structural forms, electrodes of all kinds to be used in current production, metal deposition, and chemicals manufacturing [92]. Their fabrication involves a number of specific operations and processes. For instance, carbon blacks are deposited, collected, and processed. Cokes must be crushed and calcined; binders (pitches) must be pulverized and classified. Green mixtures are formed, molded, extruded, baked, and some carbons are also graphitized to provide special properties.
References
507
Different electrode designs were developed. Porous conductive electrodes having at least two zones can be used either as reversed dual porosity electrode or as electrode assembly with conductive, noncompressible porous carbon matrices [92]. The gas-diffusion electrode constitutes a system in which a reactive gas is supplied under pressure to a porous electrode partition that separates gas and electrolyte phases from each other [93]. By adjusting the gas pressure and average pore diameter, the electrolyte fills only part of the pore chemical system. In recent years, the preparation and properties ofPt-Ru/C electrocatalysts for polymer electrolyte fuel cell applications have received considerable attention [94-97].
ACKNOWLEDGMENTS This work was financially supported by the Consejo Nacional de Investigaciones Cientificas y Tecnicas (CONICET), Agencia Nacional de Promoci6n Cientifica y Tecno16gica (PICT 98 06-03251) of Argentina, and the Comisi6n de Investigaciones Cientificas de la Provincia de Buenos Aires (CIC).
REFERENCES 1. Parsons, R. (1959). Equilibrium Properties of Electrified Interfaces, Modern Aspects of Electrochemistry, Vol. 1. Butterworth. 2. Conway, B.E. (1965). Theory and Principles of Electrode Processes. Ronald Press Co. 3. Pourbaix, M. (1966). Atlas of Electrochemical Equilibria in Aqueous Solutions. Pergamon Press. 4. de Bethune, A.J. and Swendeman Loud, N.A. (1964). The Encyclopedia of Electrochemistry. Reinhold, p. 414. 5. Charlot, G., Bezier, D., and Courtot J. (1958). Tables of Constants and Numerical Data, Vol. 8. Pergamon. 6. Randin, J .-P. (1976). Encyclopedia of Electrochemistry of the Elements, Vol. VII. Marcel Dekker, Chapter VII-l. 7. Pourbaix, M. (1970). Rapport Technique CEBELCOR, 115, RT 181-182; Science et Technique, No.7/7, 9/10, 11/12 (1947) and 1/2 (1948). 8. Pourbaix, M. (1973). Lectures on Electrochemical Corrosion. Plenum Press. 9. Delimarskii, I.K. and Markov, B.F. (1961). Electrochemistry of Fused Salts. The Sigma Press. 10. Alabyshev, A.F., Lantratov, M.F., and Morachevskii, A.G. (1965). Reference Electrodes for Fused Salts. Press Publishers, p. 112. 11. Creighton, H.J. (1951). Principles and Applications of Electrochemistry, Vol. I. John Wiley & Sons.
508
Chapter 19 Electrochemical Behavior of Carbon Materials
12. Beck, F. (1997). Graphite, Carbonaceous Materials and Organic Solids as Active Electrodes in Metal-Free Batteries, Advances in Electrochemical Science and Engineering, Vol. 5. Wiley-VCH. 13. Kinoshita, K. (1988). Carbon: Electrochemical and Physicochemical Properties. Wiley. 14. Weinberg, N.L. (1982). Techniques of Electro-Organic Synthesis, Vol. 1-3. 15. Baizer, M. and Lund, H. (1985). Organic Electrochemistry. Marcel Dekker. 16. Conway, B.E. (1999). Electrochemical Supercapacitors. Kluwer Academic Plenum Publishers. 17. Dresselhaus, G. (1984). Electronic and lattice properties of graphite. Proceedings of the Workshop on the Electrochemistry of Carbon. Pennington: The Electrochemical Society, p. 5. 18. Dresselhaus, M.S., Dresselhaus, G., and Saito, R. (1999). Nanotechnology in carbon materials. In Nanotechnology (G.L. Timp, ed.). Springer, Chapter 7. 19. Saito, R., Dresselhaus, G., and Dresselhaus M.S. (1996). Tunneling conductance of connected carbon nanotubes. Phys. Rev. B, 53, 2044-50. 20. McCreery, R.C. (1991). Structural effects of carbon electrodes. Electroanalytical Chemistry, Vol. 17. Marcel Dekker. 21. Thrower, P.A. (1982). Microstructures of carbon materials. Proceedings of the Workshop on the Electrochemistry of Carbons, Vol. 84-5. Pennington: The Electrochemical Society, p. 40. 22. Lamb, A.B. and Elder, L.W. (1931). The electromotive activation of oxygen. J. Am. Chem. Soc., 53, 137-63. 23. Singer, L.S. (1982). Electron paramagnetic resonance (EPR) in carbons. Proceedings of the Workshop on the Electrochemistry of Carbons, Vol. 84-5. Pennington: The Electrochemical Society, p. 26. 24. Golden, T.C., Jenkins, R.G., Otake, Y., and Scaroni, A.W. (1982). Oxygen complexes on carbon materials. Proceedings of the Workshop on the Electrochemistry of Carbons, Vol. 84-5. Pennington: The Electrochemical Society, p. 61. 25. Studebaker, M.L., Hoffman, E.W.D., Wolfe, A.C., and Nabors L.G. (1956). Oxygen containing groups on the surface of carbon black. Ind. Eng. Chem., 48, 162-6. 26. Randin, J.P. and Yeager, E. (1972). Differential capacitance study on the basal plane of stress-annealed pyrolytic graphite. J. Electroanal. Chem., 36, 257-76. 27. Morcos, I. and Yeager, E. (1970). Kinetic studies of the oxygen peroxide couple on pyrolytic graphite. Electrochim. Acta, 15, 953-1. 28. Yeager, E. Molla, J.A., and Gupta, S. (1982). The electrochemical properties of graphite and carbon. Proceedings of the Workshop on the Electrochemistry of Carbons, Vol. 84-5. Pennington: The Electrochemical Society, p. 123. 29. Fabish, F.J. and Schleifer, D .E. (1982). Surface chemistry and the carbon black work function. Proceedings of the Workshop on the Electrochemistry of Carbons, Vol. 84-5. Pennington: The Electrochemical Society, p. 79. 30. Aleksandrov Yu.!. and Mashovets, V.P. (1966). Wettability of the graphite electrode in molten chlorides. Zh. Prikl. Khim., 39, 2591-6. 31. Bukun, N.G. and Tkacheva, N.S. (1969). Double-layer capacitance ofa graphite electrode in molten chlorides. Elektrokhimiya, 5, 596-8. 32. Ukshe, E.A. and Leonova, L.S. (1970). Chlorine diffusion in fused lithium chloride, Elektrokhimiya, 6, 1423-5. 33. Thonstadt, J. (1970). The electrode reaction on the C, CO 2 electrode in cryolitealumina melts. II. Impedance measurements. Electrochim. Acta, 15, 1581-95.
References
5°9
34. Thonstadt, J. (1973). Double-layer capacity of graphite in cryolite-alurnnina melts and surface area changes by electrolyte consumption of graphite and baked carbon. J. Appl. Electrochem., 2, 315-19. 35. Gewirth, A.A. and Bard, AJ. (1988). In situ scanning tunneling microscopy of the anodic oxidation of highly oriented pyrolytic graphite surfaces.]. Phys. Chern., 92, 55663-6. 36. Bunde, A. and Havlin, S. (1996). Fractals and Disordered Systems. Springer. 37. Bard, AJ. and Faulkner, L.R. (1980). Electrochemical Methods. John Wiley & Sons. 38. Pyun, S.1. and Rhee, C.K. (2004) An investigation of fractal characteristics of mesoporous carbon electrodes with various pore structures. Electrochim. Acta, 49, 4171-80. 39. Ebert, L.B. (1982). Electrochemistry of intercalation compounds of graphite. Proceedings of the Workshop on the Electrochemistry of Carbons, Vol. 84-5. Pennington: The Electrochemical Society, p. 595. 40. Murray, F. (1999). Modified electrodes. Electroanalytical Chemistry, Vol. 13. Marcel Dekker. 41. Zagal,J. Sen, R.K., and Yeager, E. (1977). Oxygen reduction by Co(II) tetrasulfonatephthalocyanine irreversibly adsorbed on a stress-annealed pyrolytic graphite electrode surface.]. Electroanal. Chem., 83, 207-13. 42. Brown, A.P. and Anson, F.C. (1977). Molecular anchors for the attachment of metal complexes to graphite electrode surfaces. J. Electroanal. Chem., 83, 203-7. 43. Schreurs, J. and Barendrecht, E. (1984). Surface-modified electrodes. Reel. Trav. Chim. Pays-Bas, 103,205-19. 44. Dryhurst, G. (1977). Electrochemistry if Biological Molecules. Academic Press. 45. Matsue, T. FUjihira, M., and Osa, T. (1979). Selective chlorination with a cyclodextrin modified electrode.]. Electrochem. Soc., 126, 500-1. 46. Tarasevich, M.R., Yaropolov, A.I., Bogdanoskaya, V.A., and Varfolomeev S.D. (1979). Electrocatalysis of a cathodic oxygen reduction by laccase.]. Electroanal. Chern., 104, 393-403. 47. Yamaguchi, N.O., Nishiki, Y., Tokuda, K., and Matsuda, H. (1982). Apparent diffusion coefficients for electroactive anions in coatings of protonated poly(4-vinylpiridine) on graphite electrodes.]. Electroanal. Chem., 139,371-82. 48. Wightman, R.M., Kovach, P.M., Kuhr, W.G., and Stutts, KJ. (1982). Increasing electrochemical reversibility at carbon electrodes. Proceedings of the Workshop on the Electrochemistry of Carbons, Vol. 84-5. Pennington: The Electrochemical Society, p.510. 49. McLean, J.D. (1982). Carbon electrodes for liquid chromatography detection. Anal. Chem., 54, 1169-74. 50. Anderson, J.E. Hopkins, D. Shadrick, J.W., and Ren Y. (1989). Apparatus for the fabrication ofpoly(chlorotrifluoroethylene) composite electrodes. Anal. Chem., 61, 2330-2. 51. Park, J. and Shaw, B.R. (1989). Electrochemical performance of crosslinked poly(styrene)-co-poly(vinylpyridine) composite electrodes containing carbon blacks. Anal. Chern., 61, 848-52. 52. Falat, L. and Cheng, H.Y. (1983). Electrocatalysis of ascorbate and NADH at a surface modified graphite epoxy electrode.]. Electroanal. Chem., 157, 393-7. 53. de Bettencourt-Dias, A., Winkler, K., Fawcett, W.R., and Balch, A.L. (2003). The influence of electroactive solutes on the properties of electrochemically formed fullerene C 6o -based films.]. Electroanal. Chem., 549, 109-17.
51 0
Chapter 19 Electrochemical Behavior of Carbon Materials
54. Tan, W.T., Bond, A.M., Ngooi, S.W., et al. (2003). Electrochemical oxidation of L-cysteine mediated by a fullerene-C 6o -modified carbon electrode. Anal. Chim. Acta, 491, 181-91. 55. Levich, B.G. (1962). Physicochemical Hydrodynamics. Prentice-Hall. 56. Albery, W.J. and Hitchmann, M.1. (1971). Ring-Disc Electrodes. Clarendon Press. 57. Zagal, J., Bindra, P., and Yeager, E. (1980). A mechanistic study of oxygen reduction on water soluble phtalocyanine adsorbed on graphite electrodes.]. Electrochem. Soc., 127, 1506-17. 58. Collman, J.P., Denisevich, P., Konai, K., et al. (1980). Electrode catalysis of the four-electron reduction of oxygen to water by dicobalt face-to-face porphyrins. ]. Am. Chem. Soc., 102, 6027-36. 59. Damjanovic, A., Genshaw, M.A., and Bockris, J.O'M. (1967). The role of hydrogen peroxide in oxygen reduction at platinum in H 2 S0 4 solution.]. Electrochem. Soc., 114,466-72. 60. Lu, J.T., Tryk, D., and Yeager, E. (1982). Determination of the equilibrium constant for the superoxide dismutation. Extended Abstracts of the Electrochemical Society Meeting, Abstract 82-1. 61. Gimeno, Y., Hernandez Creus, A., Gonzalez, S., et al. (2001). Preparation of 100-160 nm sized branched Pd islands with enhanced electrocatalytic properties in HOPG. Chem. Mater., 13, 1857-64. 62. Gimeno, Y., Hernandez Creus, A., Carro, P., et al. (2002). Electrochemical formation of Pd islands on HOPG: kinetics, morphology and growth mechanism. ]. Phys. Chem. B, 106, 4232-44. 63. Boxley, Ch., J., White, H.S., Listex, T.E., and Pinhero, P.J. (2003). Electrochemical deposition and reoxidation of Au at HOPG. Stabilisation of Au nanoparticles on the upper plane of step edges.]. Phys. Chem. B, 107,451-8. 64. Liu, H.Y., Weaver, MJ., Wang, C.B., and Chang C.K. (1983). Dependence of electrocatalysis for oxygen reduction by adsorbed dicobalt cofacial porphyrins upon catalyst structure.]. Electroanal. Chem., 145, 439-47. 65. Iliev, I. (1981). Air cathodes for primary metal air batteries. Extended Abstracts of the Electrochemical Society National Meeting. 66. Salimi, A., Banks, C.E., and Compton, R.G. (2003). Ultrasonic effects on the electroreduction of oxygen at a glassy carbon anthraquinone-modified electrode. The Koutecky-Levich equation applied to insonated electrocatalytic reactions. Phys. Chem. Chem. Phys., 5, 3988-93. 67. Szucs, A., Budavari, V., Berkesi, 0., and Novak M. (2003). Electrochemical hydrogenation of C 60 fullerene films.]. Electroanal. Chem., 548, 131-7. 68. Tarasevich, M.R. and Khrushcheva, E.1. (1989). Electrocatalytic Properties if Carbon Materials, Modern Aspects if Electrochemistry, Vol. 19. Plenum Press, p. 2J5. 69. Janssen, L.J. and Hoogland, J.G. (1970). The electrolysis of an acidic NaCl solution with a graphite anode. III. Mechanism of chlorine evolution. Electrochun. Acta, 15, 941-51. 70. Fang, C. and Zhon, Y. (2001). The electrochemical characteristics of C 6o -glutathione modified Au electrode and the electrocatalytic oxidation of NADH, Electroanalysis, 13, 949-54. 71. Chen, T., Barton, S., Binyamin, G., et al. (2001). A miniature biofuel cell.]. Am. Chern. Soc., 123,8630-1. 72. Mano, N., Mao, F., and Heller, A. (2002). A miniature biofuel cell operating in a physiological buffer.]. Am. Chem. Soc., 124, 12962-5.
References
511
73. Stonehart, P. and MacDonald,].P. (1982). Corrosion of carbons in acid electrolytes. Proceedings of the Workshop on the Electrochemistry of Carbons, Vol. 84-5. Pennington: The Electrochemical Society, p. 292. 74. Kokhenov, G. and Milova, N. (1969). Effect of pH on the anodic oxidation of graphite. Elektrokhimiya, 5, 93-6. 75. Kinoshita, K. and Bett,]. (1973). Electrochemical oxidation of carbon black in concentrated phosphoric acid at 135C. Carbon, 11, 237. 76. Kinoshita, K. and Bett, J.A.S. (1974). Corrosion problems in energy conversion and generation. The Electrochemical Society. 77. Koehler, W.A. (1951). Principles and Applications of Electrochemistry, Vol. II. John Wiley & Sons. 78. Galloni, P. (1973). Trattato di Ingenieria Eletrochimica. Tamburini. 79. Thonstadt, J. (1970). The electrode reaction on the C, CO 2 electrode in cryolitealumina melts. I. Steady state measurements. Electrochim. Acta, 15 1569-80. 80. Arvia, A.J. and de Cusminsky, J.B. (1967). EI Potencial del Electrodo de Fluor en el Bifluoruro de Potasio Fundido. An. Asoc. Quim. Arg., 55, 41-6. 81. Arvia, A.J. and de Cusminsky, J.B. (1962). Kinetics of the electrochemical formation of fluorine at carbon electrodes. Trans. Faraday Soc., 58, 1019-32. 82. Vandenbroele, H.J. and Arvia, A.J. (1967). Estudio Cinetico del Electrodo de Cloro en Medios I6nicos Fundidos. An. Asoc. Quim. Atg., 55, 21-40. 83. Triaca, W.E. Solomons, C., and Bockris, J.O.M. (1968). The mechanism of the electrolytic evolution and dissolution of chlorine on graphite. Electrochim. Acta, 13, 1949-64. 84. Arvia, A.J. and Triaca, W.E. (1965). Anodic reactions of molten nitrates on graphite. Electrochim. Acta, 10, 1188-9. 85. Arvia, A.J. and Triaca, W.E. (1966). Electrolysis of molten nitrates on graphite electrodes: kinetics of the anodic reaction. Electrochim. Acta, 11, 975-88. 86. Arvia, A.J. and Calandra, A.J. (1967). Kinetics of the discharge of nitrite ions in the electrolysis of molten nitrites on graphite electrodes. Electrochim. Acta, 12, 1441-55. 87. Sustersic, M.G. Triaca, W.E., and Arvia, A.J. (1974). Potentiodynamic behaviour of graphite and platinum electrodes in sodium nitrite-potassium nitrite melts. Electrochim. Acta, 19, 19-25. 88. Balskus, E.J., Podesta, J.J., and Arvia, A.J. (1970). Kinetics of electrochemical hydrogen evolution and dissolution on graphite in molten KHS0 4 . Electrochim. Acta, 15, 1557-8. 89. Balskus, E.J., Podesta, J.J., and Arvia, A.J. (1971). Hydrogen evolution and dissolution on graphite electrodes in the electrolysis of molten KHSO 4' I. Kinetics of the reactions on dense graphite. Electrochim. Acta, 16, 1663-70. 90. Balskus, EJ., Triaca, W.E., and Arvia, AJ. (1972). Hydrogen evolution and dissolution on graphite electrodes in molten potassium bisulphate. II. Kinetics and mechanism of the reactions on porous graphite. Electrochim. Acta, 17, 45-62. 91. Arvia, A.J., Triaca, W.E., and Videla, A.H. (1970). Kinetics and mechanism of the electrochemical oxidation of graphite in bisulphate melts. Electrochim. Acta, 15,9-24. 92. Kordesch, K., Jahangir, S., and Schautz, M. (1982). Carbon electrodes manufacturing techniques. Proceedings of the Workshop on the Electrochemistry of Carbons, Vol. 84-5. Pennington: The Electrochemical Society, p. 387. 93. Bockris, J.O'M. and Srinivasan, S. (1969). Fuel Cells: Their Electrochemistry. McGraw-Hill.
512
Chapter 19 Electrochemical Behavior of Carbon Materials
94. Schneider, J., Wambach, C., Pennemann, B., and Wandelt, K. (1999). Scanning tunneling microscopy and scanning tunneling spectroscopy studies of powdery palladium/graphite model catalysts. Langmuir, 15, 5765-72. 95. Pozzio, A., Silva, R.F., De Francesco, M., et al. (2002). A novel route to prepare stable PtRu/C electrocatalysts for polymer electrolyte fuel cell. Electrochim. Acta, 48,255-62. 96. Steigerwalt, E.S., Deluga, G.A., Cliffel, D.E., and Lukehart, C.M. (2001). A Pt-Ru/graphitic carbon nanofiber nanocomposite exhibiting high relative performance as a direct-methanol fuel cell anode catalyst. J. Phys. Chem. B, 105, 8097-101. 97. Adora, S., Soldo-Olivier, Y., Faure, R., et al. (2001). Electrochemical preparation of platinum nanocrystallites on activated carbon studied by X-ray absorption spectroscopy. J. Phys. Chem. B, 105, 10489-95.
SELF-AsSEMBLED MONOLAYERS ON (0001) Fernando Teran Arce,* Jose L. Zubimendi, Maria E. Vela, Roberto C. Salvarezza, and Alejandro J. Arvia Instituto de Investigaciones Fisicoqufmicas Te6ricas y Ap/icadas. (lNIFTA), Universidad Nacional de La Plata-Consejo Nacional de Investigaciones Cientfficas y Tecnicas, La Plata, Argentina *Present address: Center for Nanomedicine, Department of Medicine, University of Chicago, Chicago, IL, USA
Contents Introduction Characteristic of the HOPG Substrate 20.3 Self-Assembled Submonolayers and Monolayers Acknowledgments References 20.1
51 3
20.2
51 4 521 52 7 52 7
20.1 INTRODUCTION Scanning nanoscopies have led to a new stage in the study of interfacial processes. Data derived from these techniques, especially scanning tunneling microscopy (STM) and atomic force microscopy (AFM), offer the possibility of studying the physical chemistry of surfaces on solid substrates at the atomic and molecular level [1-5]. Heterogeneous catalysis is an important field for the application of these techniques. Because of the use of these nanoscopies, advances have been made in the knowledge of the geometry and effective area of solid catalysts, the sintering process that decreases their performance and lifetime, the adsorbate film structure on crystallographically well-defined surfaces, and the influence of surface defects on the dynamic behavior of these films during adsorption, desorption, and chemical reaction stages. Adsorption by Carbons ISBN: 978-0-08-044464-2
© 2008 Elsevier Ltd. All rights reserved.
513
Chapter 20 Self-Assembled Monolayers on C(OOOl)
Nanometer-scale (nm) nanoscopies provide information on restricted molecular domains that comprise some hundreds of molecules. This information at the local level is not accessible by other surface analysis techniques because the latter provide average data on the whole sample. Studies at the local level reveal the complexity of physicochemical processes taking place at solid/fluid interfaces under different perturbation conditions. Local data are a solid basis for the theoretical interpretation of these processes by the use of Quantum Mechanics procedures. Nanoscopies supplemented with conventional techniques will allow the rational handling of the catalyst/reactive system based on its knowledge at the atomic/molecular level. The application of nanoscopies in surface chemistry offers the possibility for determining the nanostructure of solid surfaces, surface reconstruction phenomena, to identify the structure of ionic and molecular adlayers, to study the dynamics of these adlayers in their adsorption and desorption at the submonolayer and monolayer (ML) level. Likewise, they are important tools to follow reactions at solid surfaces in real time in different environments. The reader can get acquainted with the state of the art on these topics in Refs [5-12]. This chapter describes the application of tunneling and AFM to the study of inorganic and organic adsorbates on C(OOOl) at the submonolayer and ML level. The C(OOOl) surface can be taken as a model system for the study of adsorption processes because it is atomically smooth and exhibits a low chemical reactivity, allowing an easy handling in the atmosphere. The knowledge of adsorption on carbon is important in the field of electrocatalysis because carbon is widely used as a matrix for the dispersion of catalytically active metallic clusters.
20.2 CHARACTERISTIC OF THE 20.2.1
HOPG
SUBSTRATE
General Considerations
Highly oriented pyrolytic graphite (HOPG) is the most adequate type of carbon to investigate the adsorption of both molecules and atoms, and the formation of molecular and atom clusters on the C(OOOl), the basal plane of graphite [1-5],[9-12] The procedure for HOPG fabrication was developed by Union Carbide in USA. HOPG is prepared from the thermal decomposition of gaseous hydrocarbons on a surface heated at 1200-3800°C followed by highpressure compression of the surface under heating [13]. HOPG's first use was as an X-ray diffraction grating. Later, with the advent of nanoscopies it became of particular interest as a carbon material with flat terraces constituted by the basal plane that could be resolved at the atomic scale. Accordingly, HOPG was utilized as a calibration standard for STM, and as a substrate for adsorption studies. Besides, the ease with which HOPG can have a
20.2
Characteristic of the HOPG Substrate
pristine basal plane surface just by exfoliation with scotch tape and its chemical inertia make HOPG a very important substrate to be utilized in STM and AFM. HOPG consists of ordered layers (graphene sheets) of carbon atoms constituting a honeycomb lattice. The arrangement of graphene sheets is of the type A.B.A.B (Fig. 20.1 (a)), the nearest neighbor graphene sheets are shifted horizontally by one interatomic distance [14]. The separation distance of nearest neighbor graphene sheets is 0.355 nm, and the lattice constant in the vertical direction is 0.67 nm. Correspondingly, for alternatively located graphene sheets, three carbon atoms out of the six atoms forming each hexagon of the 2D lattice lie on the same vertical, whereas the remaining three carbon atoms lie on the vertical containing the center of hexagons (Fig. 20.1 (b)). For each graphene sheet the atomic lattice consists ofsix carbon atoms forming an open honeycomb type hexagon with 0.142 nm between nearest neighbor atoms. The Bravais lattice, however, corresponds to a hexagonal lattice centered with two carbon atoms for each unit cell, and 0.246 nm separation between neighbor Bravais point so that 0.246/0.142 = J3. The interaction between graphene layers is determined by van der Waals weak forces, making exfoliation of HOPG easy. Graphite is thermodynamically stable under usual conditions, but its structure is typically anisotropic as it is reflected, for instance, by the Young's modulus that is 10.3 x 10- 5 MPa along the basal plane and 0.3 x 10- 5 MPa in the direction perpendicular to the basal plane. A similar effect occurs with the capacitance of the HOPG/aqueous electrolyte interface, the potential of zero charge, and the work function values (see Chapter 21).
1,42A!
(a)
(b)
Figure 20.1 (a) Scheme of the 3D higWy oriented pyrolytic graphite (HOPG) atomic layer. Note the lateral displacement of atomic layers. (b) Little circles form the honeycomb lattice. Big circles correspond to the Bravais cell. The unit cell is drawn (shadow) at the upper right part of the figure. The corrugation between two carbon atoms located within the ellipse is the region sensed by the tip in the contact mode atomic force microscopy (AFM).
Chapter 20 Self-Assembled Monolayers on C(OOOl)
51 6
The four valence electrons of carbon are involved in three IT bonds and one 'IT bond with its neighbors in plane. The electrical conductivity of graphite is due to 'IT bonded electrons. In contrast to insulator diamond, the electrical resistance of graphite along the basal plane direction is 4.1 x 10- 5 n cm, a figure of the same order of magnitude as that of metals such as platinum and palladium. According to the band theory, graphite is considered as a semimetal, the overlapping of the conduction and valence bands is about 0.04 eVe The electronic structure of graphite accounts for its hydrophobicity [15]. 20.2.2
Nanoscopy Characterization of HOPG
20.2.2.1
The hexagonal lattice
AFM images (10 x 10 f-Lm 2 ) of a fresh HOPG surface (Fig. 20.2) show a number of features, namely, large monoatomic terraces about 100 nm wide and several micrometer long. Terraces are separated by steps of either one atom or a few atoms in height. Some triangular-shaped terraces with angles that are multiples of 30° are consistent with the hexagonal lattice (Fig. 20.3(a)). Besides, STM images also show some features that are artifacts from the exfoliation technique (Fig. 20. 3(b)). These artifacts have been classified as steps, strings, fibers [16] either single or agglomerated, small pieces of graphite, and very tiny particles. A detailed analysis of these additional features is required to avoid a wrong interpretation of the structure of adsorbate patterns on HOPG. When terraces are imaged at high resolution (Fig. 20.4(a)), i.e., below 10 x 10 nm2 , the STM image ofC(0001) depicts a hexagonal lattice with nearest neighbor distance d = 0.246 nm. The corrugation of this type of image depends
10.0
7.5
5.0
·2.5
o
o 2.5
5.0
7.5
10.0 Jlrn
Figure 20.2 Ex situ atomic force microscopy (AFM) image of the basal plane of highly oriented pyrolytic graphite (HOPG). Wide terraces separated by steps can be seen.
20.2
Characteristic of the HOPG Substrate
(a)
(b)
Figure 20.3 (a) Ex situ 3.8x3.8 f.1m 2 atomic force microscopy (AFM) image of highly oriented pyrolytic graphite (HOPG) that shows steps of different heights. (b) A 4.15 x 4.15 f.1m 2 AFM image of HOPG where strings produced by the exfoliation technique are shown.
(a)
(b)
Figure 20.4 (a) Ex situ 3 x 3 nm2 scanning tunneling microscopy (STM) image usually found in highly oriented pyrolytic graphite (HOPG). (b) Honeycomb structure observed by STM (1x1 nm2 ).
518
Chapter 20 Self-Assembled Monolayers on C(OOOl)
on the tunneling current (It) and the voltage (~) applied between the STM tip and the sample surface. Thus, corrugation of about 0.1 nm results for It ~ 1 nA and ~ ~ 0.05 V, whereas the corrugation decreases to 0.02 nm for ~ ~ 1 V. The change in voltage polarity has practically no effect on the HOPG image. Occasionally, the typical honeycomb structure of graphite can be observed by STM (Fig. 20.4(b)). As discussed below, the origin of this type of images is controversial. They have been considered as "high-resolution images" that are obtained when the STM tip is extremely sharp or as an "artifact" arising from a multiple tip [17]. The lattice shown in Fig. 20.1(b), which is usually imaged by STM or AFM, is formed by only three instead of six carbon atoms. The corresponding nearest neighbor carbon-carbon atom distance is that of the Bravais hexagonal lattice referred to above. The scheme depicted in Fig. 20.1 (a, b) accounts for the appearance of this image.
20.2.2.2
Additional features
Step corrugations of about 10-20 nm are easily observed with thin graphite samples that have been exfoliated several times. However, the Bravais lattice can be observed by small-size imaging (~10 x 10 nm2 ) , and at slightly higher magnifications (50 x 50 nm2 ) superstructures of different periodicity are occasionally observed. These superstructures of HOPG (Fig. 20.5.) make it difficult to recognize unambiguously the structure of molecular adsorbates.
Figure 20.5 Ex situ 46 x 46 nm2 atomic force microscopy (AFM) image of higWy oriented pyrolytic graphite (HOPG) showing a lattice that is not correspondent with the lattice shown in Fig. 20.1.
20.2
Characteristic of the HOPG Substrate
Strings are thin graphite stripes that are removed by exfoliation from steps and attached to another step [16]. The three threads shown in Fig. 20.3(b) that covered the entire image are 66 nm large and about 2.8 nm wide. Fibers are observed by STM as thin tubes about 2.5 nm in diameter and 20 nm long formed by agglomeration of threads. Atomic resolution at fibers can also be obtained, although with a poor definition. They are produced by step rupture by exfoliation. Island-like pieces, most of them at the border of holes, are also sometimes produced by exfoliation (Fig. 20.6). The islands depicted in this figure are about 0.3 nm high, a figure that is similar to the depth of holes. Different superlattices with x periodicity have been imaged. This periodicity has been related to rotation of graphite lattice [17]. These superlattices can be produced by either a multiple tip effect [17b] or electronic perturbations caused by adsorbed molecules [17c]. A hexagonal superlattice with a 4.4nm periodicity, rotated 30° with respect to the HOPG lattice, and 0.38 nm corrugation has also been reported [17a]. This superlattice was also attributed to rotation of the surface layer of graphite. As this type of superstructures is most frequently observed for thin layers of material, they have been associated with charge density waves [14, 18]. Occasionally, a sort of lattice of holes is also imaged. The structure of this lattice can be interpreted as an atomic honeycomb lattice in which each hole in the image would represent the hole of a hexagon in the honeycomb lattice. According to theoretical calculations, graphite STM images with atomic
-J3 -J3
2.00
1.00
o
o 1.00
2.00
flm Figure 20.6 Ex situ atomic force microscopy (AFM) image that shows island-like pieces produced on higWy oriented pyrolytic graphite (HOPG) by step rupture from the exfoliation procedure.
Chapter
52 0
20
Self-Assembled Monolayers on C(OOOl)
resolution should be dominated by independent Fourier components [19] of three carbon atoms usually imaged. The multiple tip effect would produce a relative change in the amplitude and phase of other components, this fact being reflected in the change of the maximum amplitude observed by STM. A poor instrument resolution might produce a comparable effect [20]. The same features from atomic resolution AFM images of graphite (Fig. 20.7(a, b)) can be distinguished. Two models have been proposed to explain the AFM images of graphite [21]. In one of these models the calculations are based on the scanning of the graphite surface with a single potassium atom. For applied forces of the order ofl nN, i.e., a value lower than about 50 nN used in the contact mode AFM, the corrugation between two carbon atoms located within the ellipse (Fig. 20.1(b)) would be indistinguishable by the AFM cantilever tip. But the situation would be reversed when the tip goes through two ellipses via the hexagon centers. Another possibility considers the asymmetry of carbon sites in the graphite lattice (Fig. 20.1 (a)). Thus, the carbon atom located in the upper graphene, which is directly above the carbon atom in the lower graphene, would suffer a weaker interaction with the tip than that facing the centers of the hexagon. This explanation would be similar to that admitted for the interpretation of the corresponding STM images. Therefore, it can be concluded that STM imaging on HOPG is influenced by structural defects, adsorbates and electronic effects [22]. The latter would prevail at step sites where an asymmetric distribution of electric charges would be more favorable.
o (a)
6.00nmO (b)
2.93nm
Figure 20_7 Ex situ atomic force microscopy (AFM) images of higWy oriented pyrolytic graphite (HOPG). The distance is 0.246 nm.
20.3
Self-Assembled Submonolayers and Monolayers
521
20.3 SELF-AsSEMBLED SUBMONOLAYERS AND MONOLAYERS
One important aspect of heterogeneous chemical reactions at solid surfaces is related to the presence of adsorbed species that play a key role in determining the rate and efficiency of these processes. Therefore, the knowledge of molecular arrangements on solid catalysts of reactants, reaction intermediates, or products is of outstanding importance in dealing with fundamental aspects of heterogeneous catalysis. Self-assembled molecular arrangements on HOPG can be spontaneously produced by different procedures that are based on the type of interactions between either bare HOPG regions or functional oxygen-containing groups existing at the HOPG surface (see Chapter 21). These arrangements, covering from the submonolayer to the multilayer level, are dominated by either physical or chemical adsorption. The adsorption of an atom on a molecule at a solid surface is attributed to a physisorption phenomenon principally because of van der Waals forces. In physisorption no appreciable reordering in the adsorbate electronic distribution occurs. This situation is generally found in the adsorption of noble gases on metal surfaces, in which adsorption energy values in the range 1-10 k] I mol are involved. These figures are of the same order of magnitude as that of the thermal energy (k7) ofmolecules at T = 298 K (approximately 2.5 k]lmol). Accordingly, to observe physisorbed systems by atomic resolution STM or AFM, experiments have to be performed at a very low temperature and above 1 atm pressure [23]. The formation of supramolecular layers (see Chapter 21) is another way of producing adequate architectural molecular designs on HOPG and carbons in general, although the structural analysis of these layers by nanoscopic techniques is still a complicated matter. A few typical examples of adsorbates on C(OOOl) are described in the following sections. 20.3.1 Alkane Adsorption on ((0001)
In principle, the adsorption of alkane molecules on C(OOOl) would appear unlikely because of the inert character of the substrate. In this case, however, besides van der Waals forces, other contributions come into play and can make energy adsorption reach values of up to 100 kJ/mol, which are comparable to those of chemisorption processes. This enables determining the structure of aliphatic hydrocarbons adsorbed on C(OOOl) by AFM or STM because the adsorbate withstands tip-sample interaction forces. The adsorption energy of alkanes on C(OOOl) decreases with temperature and increases with the chain length [24] due to an increase in the affinity of alkane carbon atoms with C(OOOl) atoms. This involves the adsorption of aliphatic molecules ordered with the chain axis lying parallel to the C(OOOl) plane. Under these conditions, the interaction of the adsorbed molecule increases because of
522
Chapter
20
Self-Assembled Monolayers on C(OOOl)
the geometric matching of the carbon lattice of the C(OOOl) plane with that of the zigzag aliphatic chain, each CH2 occupying the hexagon area in the graphite lattice. With this configuration the adsorption energies are 21.6 k]/mol for n-hexane and 105.6 kJ/mol for n-hexadecane.
20.3.2
Sulfur Atom Submonolayers on HOPG
Sulfur electroadsorbs on HOPG from SH- -containing neutral buffered aqueous solution (pH 8) at potentials (E) close to -0.8 V (versus NHE) , i.e., at values ofE more negative than the reversible potential (Er ) for the SH- = S + H+ + ereaction. The surface coverage by sulfur atoms, estimated from the electroadsorption/electrodesorption charge, is close to 1/2. Different structures of sulfur atom submonolayers on HOPG have been observed by STM [25a, b]. One of these structures corresponds to sulfur trimers with d = 0.24 nm and S atoms atop C atoms (Fig. 20.8(a)). Conversely, for E > E r , other submonolayer structures are formed, namely, a J3J3 R30° structure with d = 0.42 nm, a sulfur atom honeycomb lattice with d = 0.24 nm, rectangular arrays of sulfur atoms with d = 0.21 nm (Fig. 20.8(b)). The influence of the HOPG surface on sulfur atom electroadsorption is reflected in the values of d = 0.42 nm and d = 0.24 nm, whereas the S-S distance, d = 0.21 nm, which is observed for E > E r , is close to that found for polysulfide species [26]. A similar behavior has been observed for sulfur atom electroadsorption on Au(lll) surfaces [26]. Adsorption energy values for S atom adsorption on HOPG in the range 30-40 kJ/mol have been evaluated theoretically [25b].
20.3.3
Alkanethiol Adsorption on ((0001)
The contrast of organic molecules adsorbed on C(OOOl) in STM images depends on the functional group at the molecule head [27]. Contrast is generally enhanced for functional groups than for aliphatic chains. For functional groups it decreases in the order SH> I>Br> NH 2. This sequence offers the possibility to discriminate the functional group from the rest of the molecule by STM. It should be noted that for OH and chloride groups, contrast is comparable to that of the remaining aliphatic chain, which turns their distinction by STM practically impossible. The structures of the CH3(CH2)22SH adlayers on C(OOOl) [28] are similar to those of the alkanes. They consist of molecular domains lying parallel to each other forming a 90° angle with respect to the chain direction. A kind of disorder is also observed in the vicinity of neighbor SH groups. Ex situ AFM images of a 1-dodecanethiol ML on C(OOOl) (Fig. 20.9(a)) exhibit an array of parallel-oriented bright rows [29]. At a higher resolution (Fig. 20.9(b)) pale bands between rows, corresponding to aliphatic chains, and bright circles along each row, attributed to S heads, can be seen. Similar images
20.3
Self-Assembled Submonolayers and Monolayers
523
(b)
Figure 20.8 Ex situ atomic resolution scanning tunneling microscopy (STM) images of sulfur atoms adsorbed on higWy oriented pyrolytic graphite (HOPG): (a) 3 x3 nm 2 ; (b) 6.32x6.32 nm2 •
are obtained by STM (Fig. 20.10) although, in this case, pale bands cannot be seen. Bands 1.2 nm in length become somewhat shorter than that of the extended chain molecule. The interband separation, which would be related to the intermolecular separation, is 0.65 nm. The angle between a row of S heads and the chain direction is 120 The S heads along a row are generally placed behind the chain of the neighbor row. 0
•
Chapter 20 Self-Assembled Monolayers on C(0001)
52 4
o
16.1 nmO
8.00 nm
(b)
(a)
Figure 20.9 Atomic force microscopy (AFM) images of l-dodecanethiol monolayer adsorbed on C(OOOl). (a) Bright spots are attributable to sulfur heads. Image (a) exhibits an array of parallel-oriented bright rows. At a higher resolution (b) pale bands between rows corresponding to aliphatic chains and bright circles along each row attributed to sulfur heads can be seen.
o (a)
40.2nm 0
20.1 nm
(b)
Figure 20.10 Scanning tunneling microscopy (STM) images of l-dodecanethiol monolayer adsorbed on C(OOOl). (a) and (b) Sulfur heads exhibit an array of parallel-oriented bright rows, At a higher resolution only the higWy oriented pyrolytic graphite (HOPG) lattice can be observed (not shown).
20.3
Self-Assembled Submonolayers and Monolayers
o (a)
7.29nm
16.1 nmO (b)
Figure
20.11 Molecular resolution atomic force microscopy (AFM) images of a l-butanethiol monolayer on higWy oriented pyrolytic graphite (HOPG). Distance between bright lines are compatible with the length of adsorbed molecules.
Ex situ AFM images of 1-butanethiol on C(OOOl) (Fig. 20.11) exhibit a structure similar to that described above. In the lower left part of the image depicted in Fig. 20.11 some bright spots along the rows, probably related to S heads, and a few pale bands, associated with aliphatic chains, can also be observed. The intermolecular separation distance between two bright spots is 0.45 nm. The 0.55-nm-Iong band is consistent with the aliphatic chain length (Fig. 20.11). As observed for 1-dodecanethiol, the angle between the chains and the direction of a bright row of S heads is 115 Alkanethiols with short- and medium-length aliphatic chains adsorbed on C(OOOl) display heads with the molecular axis lying parallel to the basal plane of the substrate. This conclusion that was drawn from the analysis of AFM images is consistent with the interrow separation ofS heads deduced from STM images, and agrees with previous results for alkanethiol with 22 carbon atoms [27, 28]. However, aliphatic chains of intermediate length seem to be extended on the substrate surface only a fraction of their length. As in the case of Au(lll), for alkanethiols adsorbed on C (000 1) the longer the adsorbate aliphatic chains the more ordered they are [30]. The fact that in some regions in the images (Figs 20.9 and 20.10) the length of 1-dodecanethiol does not match exactly that of the molecule fully extended on the surface is attributed to the occurrence of a mixed cis-trans configuration (gauche conformation). The scheme of this configuration (Fig. 20.12) includes C atoms located at the same sites as those of the C(OOOl) lattice, and the change in configuration is shown by an ellipse (see Fig. 20.12). The S atom separation between two aliphatic chains in a row is 1.25 nm, a figure that agrees with that determined from the images. The S atom separation between two molecules 0
•
526
Chapter
20
Self-Assembled Monolayers on C(OOOl)
120
a
~
W
Figure 20.12 Scheme ofthe l-dodecanethiol structure adsorbed on highly oriented pyrolytic graphite (HOPG). The axis of the hydrocarbon chain is oriented parallel to the surface, although partially extended.
located at neighbor rows is 0.75 nm, and the angle formed between the molecule axis and the direction of S heads is 120 These figures agree reasonably well with measured values. The adsorption energy of alkanethiols on C(OOOl) can be estimated considering the energy of lateral interactions between the aliphatic chains, which is of the order of 4 kJ/mol for CH 2 , and the intermolecular interaction energy in liquid alkanes, which for hexadecane is 57 kJ/mol. The energy difference for the interaction between the chains of the hexadecanethiol ML on Au(lll) and in liquid alkane is 17 kJ/mol. From the calculation of the adsorption energy of S on C(OOOl) [31], it was concluded that the vertices of graphite hexagons (small circles in Fig. 20.1) or sites located between two hexagons (large circles in Fig. 20.1) are the most favorable adsorption sites, as shown by 1-butanethiol adsorption on C(OOOl). Predictions, however, become more uncertain because in these cases the tendency of C atoms to follow the C(OOOl) lattice prevails. On the basis of the adsorption energies of n-hexadecane (100 kJ I mol), n-hexane (22 kJ/mol) , and S on C(OOOl), the adsorption energy for dodecanethiol obtained from extrapolation is 70 kJ/mol. This value exceeds that ofthe adsorption energy of S on C(OOOl) and confirms the stability of the adsorbate on C(OOOl). The ordering of adlayers of alkanethiols on C(OOOl) indicates that the head-neighbor molecule chain interactions and, to a lesser extent, that of the head-head of molecular pairs prevail. Results from the adsorption of alkanethiols on C(OOOl) as well as on Au(lll) [32] show a strong influence of the substrate on the configuration of adsorbed molecules, regardless of the length of the aliphatic chain. 0
•
References
527
ACKNOWLEDGMENTS The authors thank the Consejo Nacional de Investigaciones Cientificas y Tecnicas (CONICET) and Agencia Nacional de Promoci6n Cientifica y Tecno16gica from Argentina for their financial support (PIP 0897, PICT 99-5030, and PICT 98 N°06-03251) to the research projects on which this chapter is based. MEV is a member of the research career of CIC.
REFERENCES 1. Binnig, G. and Rohrer, H. (1982). Surface studies by scannIng tunneling microscopy. Helv. Phys. Acta, 55, 726. 2. Rohrer, H. (1989). Scanning tunneling microscopy and related methods. NATO ASI Series, Series E, Applied Sciences, No. 184, Kluwer, p. 1. 3. Forbes, R.G. (1989). Scanning tunneling microscopy and related methods. NATO ASI E Series, No. 184, Kluwer, p. 163. 4. TersoH: J. and Hamann, D.R. (1983). Theory and application for scanning tunneling microscope. Phys. Rev. Lett., 50; Tersoff J. (1989). Scanning tunneling microscopy and related methods. NATO ASI E Series, No. 184, Kluwer, p. 77. 5. Leavens, C.R. and Aers, G.C. (1989). Scanning tunneling microscopy and related methods. NATO ASI E Series, No. 184, Kluwer, p. 27. 6. Siegenthaler, H. (1992). Scanning Tunnelling Microscopy II (R. Wiesendangerand and HJ. Guntherodt, eds) . Springer-Verlag. 7. Gewirth, A.A. and Siegenthaler, H. (1995). Nanoscale probes of the solid/liquid interface. NATO ASI Series, Applied Sciences, No. 288, Kluwer. 8. Lorenz, H.J. and Plieth, W. (1998). Electrochemical nanotechnology. In situ Local Probe Techniques at Electrochemical Interfaces. Wiley-VCH. 9. Itaya, K. (1998). In situ scanning tunneling microscopy in electrolyte solutions. Progr. Surf. Sci., 58, 121-47. 10. Capella, B. and Dieder, G. (1999). Force-distance curves by atomic force microscopy. Surf. Sci. Rep., 34, 1-3. 11. Binns, G., Baker, S.H., Demangeat, C., and Parlebas,J.C. (1999). Growth, electronic, magnetic and spectroscopic properties of transition metals on graphite. Surf. Sci. Rep., 34, 107-70. 12. Binnig, G., Quate, C.F., and Gerber, C. (1986). Atomic force microscope.Phys. Rev. Lett.,56, 930-3. 13. Mc.Creery, R.L. (1994). Carbon electrodes: structural effects on electron transfer kinetics, electroanalytical chemistry, Vol. 17 (A.J. Bard, ed.). Marcel Decker, p. 221. 14. Wiesendanger, R. and Anselmetti, D. (1994). Scanning Tunneling Microscopy 1. (HJ. Giintherodt and R. Wiesendanger, eds). Springer-Verlag. 15. Adamson, A.W. (1990). Physical Chemistry of Surfaces. Wiley-Interscience. 16. Chang, H. and Bard, AJ. (1991). Observation and characterization by scanning tunneling microscopy of structures generated by cleaving highly oriented pyrolytic graphite. Langmuir, 7, 1143-53.
528
Chapter 20 Self-Assembled Monolayers on C(OOOl)
17. (a) Liu, C.-Y., Chang, H., and Bard, AJ. (1991). A large scale hexagonal domainlike structures superimposed on the atomic corrugation of a graphite surface observed by scanning tunneling microscopy. Langmuir, 7, 1138-42; (b) Albrecht, T.R., Mizes, H.A., Nogami, J., et al. (1988). Observation of tilt boundaries in graphite by scanning tunneling microscopy and associated multiple tip effects. Appl. Phys. Lett. ,52, 362-4; (c) Mizes, H.A. and Foster, J.S. (1989). Long-range electronic perturbations caused by defects using scanning tunneling microscopy. Science, 244, 559-62. 18. Coleman, R.V., Dai, Z., McNairy, W.W., et al. (1993). Methods of experimental physics. In Scanning Tunneling Microscopy. Vol. 27 G.A. Stroscio and WJ. Kaiser, eds). Academic Press. 19. Mizes, H.A., Park, S., and Harrison, W.A. (1987). Multiple-tip interpretation of anomalous scanning-tunneling-microscopy images of layered materials. Phys. Rev. B, 36, 4491-4. 20. Binnig, G., Fuchs, H., Gerber, Ch., et al. (1986). Energy-dependent state-density corrugation of a graphite surface as seen by scanning tunneling microscopy. Europhys. Lett., 1, 31-6. 21. Lin, F. and Meier, D J. (1994). Atomic-scale resolution in atomic force microscopy. Langmuir, 10, 1660-2. 22. McDermott, M.T. and McCreery, R.L. (1994). Scanning tunneling microscopy of ordered graphite and glassy carbon surfaces: electronic control of quinone adsorption. Langmuir, 10, 4307-14. 23. Somorjai, A.G. (1981). Chemistry in Two Dimensions: Surfaces. Ithaca: Cornell University Press, p. 178. 24. Findengg, G.H. (1972). Ordered layers of aliphatic alcohols and carboxylic acids at the pure liquid/graphite interface. J. Chem. Soc. Faraday Trans., 68, 1799-806; Ikai, A. (1996). STM and AFM of bioiorganic molecules and structures. Surf. Science Rep., 26, 261-332; Giancarlo, L.C. and Flynn, G.W. (1998). Scanning tunneling and atomic force microscopy probes of self-assembled, physisorbed monolayers: peeking at the peaks. Annu. Rev. Phys. Chem., 49, 297-336; Xie, Z.X., Xu, X., Mao, B.W., and Tanaka, K. (2002). Self-assembled binary monolayers of n-alkanes on reconstructed Au(111) and HOPG surfaces. Langmuir, 18, 3113-16. 25. (a) Zubimendi, J.L., Salvarezza, R.C, Vazquez, L., and Arvia, AJ. (1996). Scanning tunneling microscopy observation of sulfur electrodeposits on graphite single crystals. Langmuir, 12, 2-11; (b) Vicente, J.L. Mola, E.E., Appignanessi, G., et al. (1996). A quantum chemistry approach to possible sulfur adsorbate structures on the basal plane of graphite clusters. Langmuir, 12, 19-22. 26. Vericat, C., Andreasen, G., Vela, M.E., and Salvarezza, R.C. (2000). Dynamics of potential-dependent transformations in sulfur adlayers on Au(111) electrodes.]. Phys. Chem. B, 104, 302-7; Andreasen, G., Vericat, C., Vela, M.E., and Salvarezza, R. C. (1999) . Dynamics of sulfur adlayer transformations at metal/electrolyte interfaces.]. Chem. Phys., 111, 9457-60. 27. Venkataraman, B., Flynn, G.W., Wilbur, J.L., et al. (1995). Differentiating functional groups with the scanning tunneling microscope.]. Phys. Chem., 99, 8684-9. 28. Cyr, D.M., Venkataraman, B., Flynn, G.W., et al. (1996). Functional group identification in scanning tunneling microscopy of molecular adsorbates. J. Phys. Chem., 100, 13747-59.
References
29. Teran Arce, F., Vela, M.E., Salvarezza, R.C., and Arvia, A.J. (1996). Comparative study ofthiol films on C(OOOl) and Au(lll) surfaces by scanning probe microscopy. Surf. Rev. Lett., 4, 637-49. 30. Porter, M.D., Bright, T.B., Allara, D.L., and Chidsey, C.E.D. (1987). Spontaneously organized molecular assemblies. 4. Structural characterization of n-alkyl thiol monolayers on gold by optical ellipsometry, infrared spectroscopy, and electrochemistry. J. Am. Chem. Soc., 109, 3559-68. 31. Stranick, S.J., Parikh, A.N., Allara, D.L., and Weiss, P.S. (1994). A new mechanism for surface diffusion: motion of a substrate-adsorbate complex. J. Phys. Chem., 98, 11136-42. 32. Ulman, A. (1991). An Introduction to Ultrathin Organic Films:ftom Langmuir-Blodgett to Self-Assembly. Academic Press; Finklea H.G. (2000). Encyclopedia ofAnalytical Chemistry: Applications, Theory and Instrumentation (R.A. Meyers, ed.). Wiley; Schreiber, F. (2000). Structure and growth of self-assembling monolayers. Prog. Surf. Sci., 65. 151-257.
REMOVAL OF INORGANIC GASES AND VOCS ON ACTIVATED CARBONS Teresa
J. Bandosz
Department of Chemistry, City College
of New York,
New York, NY, USA
Contents Introduction 21.2 Adsorption of Inorganic Gases 21.3 Adsorption of Volatile Organic Compounds 21.4 Choice of Proper Carbon for a Desired Application References 21.1
533
534 549
553 556
21.1 INTRODUCTION Industrial revolution, along with development of new technologies to
improve everyday life, resulted in emission to the atmosphere vast quantities of athropogenic gases and toxic and cancerogenic volatile organic compounds (VOCs). Some of those species, as hydrogen sulfide or sulfur dioxide, have also their natural sources such as geothermal vents, volcanoes or other natural technologies where anaerobic digestion is the main bacterial activity. But it was a human addition to mother nature, which has resulted in detrimental environmental changes such as acid rains, photochemical smog, or global warming [1,2]. It is estimated that every year around 100 millions tons of 50 2 and N0 2 are emitted to the atmosphere from anthropogenic sources [1], mainly from power plants where fossil fuel is burned. The major sources of air pollution were, and still are, highly industrialized countries such as the United States or European nations. The situation changed in early 1990s when Clean Air Act of US government was introduced [2, 3]. The new regulations caused that in 2000 emissions ofacidic Adsorption by Carbons ISBN: 978-0-08-044464-2
© 2008 Elsevier Ltd.
All rights reserved.
533
Chapter
534
21
Removal of Inorganic Gases and VOCs on Activated Carbons
gases in the United States were 48 % lower than 1980 levels [3]. To complain with Clean Air Act, the new technologies engaged in desulfurization of fuel, cleaning the stock gases, or improving the efficiency of combustion have been developed and introduced. This resulted in a dramatic decrease in acidic gas emissions and significant improvements of the air quality. Nevertheless, the air quality is still controlled and the levels of pollutants such as sulfur dioxide, hydrogen sulfide, nitric dioxide, or VOCs are kept below certain thresholds considered as healthy for environment and human beings. Moreover, the stricter regulations are about to be introduced, which, for instance will require to limit the levels ofsulfur-containing species in gasoline and fuel oils to 30 and 15 ppm, respectively [4]. To follow the environmental law and to remove small but sometimes persistent concentrations of pollutants activated carbons seem to be the media of choice. They are relatively inexpensive, easily to obtain, and owing to their enormously high surface area and pore volume, they are able to remove and retain even traces of air and water pollutants. Activated carbons, due to their unique surface chemistry act not only as adsorbents but also as catalysts for oxidation of inorganic and organic species. Moreover, their surface can be modified and tailored toward desired applications. This chapter provides a comprehensive summary of surface science involved in the application of activated carbon for air cleaning from inorganic gases such as hydrogen sulfide, sulfur dioxide, nitric dioxide, hydrogen cyanide, and from VOCs. The emphasis is placed on the role of activated carbons surfaces, either unmodified or modified in the processes of adsorption and catalytic oxidationreduction of these pollutants.
21.2 ADSORPTION OF INORGANIC GASES 21.2.1
Removal of Hydrogen Sulfide
One of the leading malodorants arising from sewage treatment facilities and geothermal vents is hydrogen sulfide [5]. HzS emitted to the atmosphere is oxidized to sulfur dioxide, which results in the deposition of acid rain. Traditionally, activated carbons used for removal of high concentrations of HzS in sewage treatment plants are those impregnated with caustic materials such as NaOH or KOH [5-9]. Air currents around odor-generating facilities are initially washed in scrubbers, during which they intake high levels of humidity, and are then blown through the activated carbon vessels [7, 8]. The residual HzS quickly reacts with the strong base and is immobilized. The presence of humidity facilitates the reaction [10-12]. The carbon bed is mostly used as a support for the caustic material and storage of the oxidation products. The removal capacity of such carbon estimated using accelerated ASTM D6646-01 test [13] exceeds 0.140 g/cm3 of carbon bed. Recent study of the adsorption/oxidation mechanism on NaOH-impregnated activated carbons showed [12] that at least
21.2
Adsorption of Inorganic Gases
535
3 moles ofH2 S are adsorbed per 1 mole ofNaOH, which indicates the catalytic effect of NaOH. NaOH shifts the dissociation of hydrogen sulfide to the right increasing the content of HS- ions, which can be further oxidized either on adsorbed sulfur or on activated carbon surface. The reaction proceeds until all NaOH is consumed in the surface reaction and deposited in the form of salts, either sulfites or carbonates, and the regeneration of basic environment does not longer occurs. The shortcoming in the applications of caustic-impregnated activated carbon is the fact that impregnation decreases the ignition temperature of the carbon and poses a hazard of self-ignition [8, 9]. Another disadvantage is the oxidation of hydrogen sulfide to elemental sulfur [8, 9, 12], which cannot be removed from carbons by washing with water [14]. Moreover, the activity of caustic carbons toward H 2 S oxidation is exhausted when the caustic is consumed and the carbon pores are blocked by sulfur and sodium or potassium salts [12]. The catalytic action of NaOH-impregnated carbon can be summarized by the following reactions [12]: NaOH+H 2S
~
NaHS+H 2O
(21.1)
2NaOH+H 2S
~
Na2S+H2O
(21.2)
NaHS+0.50 2
~
S+NaOH
(21.3)
Na2S + 0.50 2 + H 2O
~
S+2NaOH
(21.4)
HS-+H 2 O
~
H 2S+OH-
(21.5)
S2- +H2O
~
HS-+OH-
(21.6)
2NaOH + H 2SO 4
~
Na2S0 4 +2H 2O
(21.7)
Both, advantages and disadvantages of caustic-impregnated carbons directed the attention of researchers toward other impregnantes, which can sustain basic
properties with less exothermic reaction in the system. An example is potassium carbonate, which was studied in details by Przepiorski and coworkers [15, 16]. According to them, hydrogen sulfide dissolves more favorably in aqueous solution of K2C0 3 than in water. H 2S, due to its small size, is able to access the small micropores as KHS (also KHC0 3 is formed), which instantly decomposes to H 2S. That H 2S located in small pores reacts with oxygen forming elemental sulfur. Important for surface catalysis is decomposition of KHC0 3 to K2C0 3 . Since oxidation of hydrogen sulfide to sulfur either in direct reaction or via dissociation to HS- and its oxidation releases significant heat, the risk of bed self-ignition still exists. This risk of self-ignition of the carbon bed along with hazardous conditions of working with high pH carbons caused that virgin (unimpregnated) activated carbons [17-49] or carbon with specific surface modifications, such as nitrogen functionality [50-57], started to be investigated as H 2S removal media. However, considerable removal capacities for hydrogen sulfide have been reported in the literature for carbons serving at temperatures around 473 K, the use of
536
Chapter
21
Removal of Inorganic Gases and VOCs on Activated Carbons
unmodified activated carbon for H 2 S removal at the ambient temperatures [40-45], is not yet common. This might be related to a relatively low capacity of virgin carbon compared to caustic-impregnated one, which for the best materials, coconut-based carbons is seven times smaller than that on the impregnated counterparts. Moreover, the mechanism on unimpregnated carbons seems to be more complicated and very detailed features of carbon surfaces play a role in adsorption and catalytic oxidation. This causes that most of the results reported so far have been based on an empirical analysis of specific types of carbon, which are sometimes difficult to reproduce [17-57]. A simple mechanism of adsorption/oxidation of hydrogen sulfide was first proposed by Hedden and coworkers [31]. According to them, dissociation of hydrogen sulfide occurs in the film of adsorbed water at the virgin carbon surface and then hydrogen sulfide ions, HS-, are oxidized by oxygen radicals to elemental sulfur. Since then many studies have been done to account for such factors as a role of water [26, 32, 34, 36, 37, 40, 48, 49], role of oxygen [18-27], autocatalysis by sulfur [27, 28], influence of pore sizes [19, 29, 33, 35, 38], role of carbon surface chemistry [41-44], the effects of ash [49, 58-60], and last but not least, speciation of surface oxidation products [41-46]. As mentioned above, the film ofwater is necessary for dissociation ofhydrogen sulfide, if pH of the surface allows it, and thus for its oxidation. It is well known that hydrophobic nature [61] of activated carbon surface is the result of high degree of aromatization and the presence of graphene-like sheets. Adsorption of water can be enhanced when functional groups containing oxygen or nitrogen (hetoreoatoms with the ability of hydrogen bonding) exist at the edges of graphene-like sheets [62, 63]. When the adsorption of hydrogen sulfide was studied at dry and wet conditions a dramatic difference in the performance of the carbon adsorbents was noticed [49]. The capacity at dry conditions is usually small and it represents mainly physical adsorption in the small pores of carbons. In some cases the presence of moisture in the air is not enough and to get the noticeable capacity prehumidification/preconditioning of samples is necessary. It was reported that on some carbons the prehumidification could improve the capacity as much as 80 times [49]. On the other hand, the amount of water adsorbed on the surface should not be too large. The studies suggested that the affinity for water adsorption should not be greater than 5 % [41, 49] to reach the maximum capacity. It is likely that, when the carbon surface becomes too hydrophilic, the small pores are filled by condensed water and the direct contact of HS- with carbon surface in the smallest pores is limited. Another factor that plays a role is the degree of carbon oxidation [41-44, 62, 63]. When more oxygen groups are present the surface becomes more acidic suppressing dissociation ofhydrogen sulfide. Although in the majority ofstudies the presence of water was found important to enhance hydrogen sulfide adsorption, Coskun and Tollefson [18] found that the presence of water at temperatures close to ambient decreases the catalytic activity of carbon surfaces. A role of oxygen in the kinetic of the H 2 S adsorption/oxidation were studied by Tollefson and coworkers [18-27], Steijns and Mars [29], and Meeyoo
21.2
Adsorption of Inorganic Gases
537
and coworkers [26]. In general, the experiments performed with low concentrations of H 2S «3 %) in wide temperature (398-473 K) and pressure ranges (230-3200 kPa). The results of Tollefson and coworkers [23] showed that optimum temperature for high H 2S conversion and low S02 production is 448 K with 0/H 2 S ration 10.05 times the stoichiometric ratio. The process was not impeded by high water vapor content, which might be related to relatively high temperature of the process, higher than boiling point of water. The rate-limiting step for catalytic oxidation reactions was defined as either adsorption of oxygen or hydrogen sulfide from the bulk phase on the activated carbon surface. The high heats of oxygen adsorption (73.8 kJ/mol) indicate its chemisorption on the surface whereas the low value of the heat of hydrogen sulfide adsorption suggests that the sorption at elevated temperature is physical in its nature [21]. Physical nature of adsorption was confirmed by Bagreev and coworkers [47] when adsorption on H 2S was studied at elevated temperature «400 K) in the absence ofair or water. At those conditions, the heat ofH2S adsorption (between 40 and 50 kJ/mol) depends only on the pore sizes. This was an indirect proof that oxygen chemisorbed on the surface or present as functional groups is not active enough active to oxidize hydrogen sulfide [47]. On the other hand, Mikhalovsky and Zaitsev [39] found using X-ray photoelectron spectroscopy (XPS) that surface oxygen-containing functional groups contribute significantly to the formation of S02 in H 2S oxidation. They suggest that at an inert atmosphere surface oxygen-containing complexes and elemental sulfur are formed during adsorption of H 2S. An interesting effect of autocatalysis by deposited sulfur was identified by Steijns and coworkers [27, 28]. Studying adsorption of hydrogen sulfide on various adsorbents, they found that deposition of sulfur at the beginning of the removal process increases the catalytic activity of the carbon. Then, when sulfur starts to block microporosity, a rapid decrease in activity follows. It was postulated that elemental sulfur is rather in the form of radical chains than S8 rings [28]. Steijns and Mars found that the catalytic activity per square meter of total surface area is approximately proportional to the amount of adsorbed sulfur. On the other hand, no evidence of sulfur autocatalysis was found by Ghosh and Tollefson [20, 21]. In all studies of hydrogen sulfide adsorption the presence of micropores is indicated as a important factor. Although the opinions about the first location of adsorbed sulfur vary [18, 19], the filling of micropores by elemental sulfur of sulfides seems to be the limiting factor of activated carbon capacity [41, 45, 46, 48]. Steijns and Mars [19] found that the strong sulfur adsorption is in carbons having pores between 0.5 and 1 nm, which is expected based on the size of sulfur chains and the overlapping of adsorption potential in pores similar in size to the adsorbate molecule. Moreover, when sulfur is adsorbed in such small pores the presence of large polymers is unlikely, and isolated adsorbed sulfur radicals are further oxidized to S02 and then S03. On such carbons, sulfuric acid is the important product of surface reaction [41]. It was also found that when the H 2S capacity of carbons is normalized to their pore volume, the similar
Chapter
21
Removal of Inorganic Gases and VOCs on Activated Carbons
values are obtained [41, 48]. This once gain indicates that the volume of pores where oxidation products are stored is the limiting factor for the amount ofH 2 S retained on the surface. A role of surface chemistry was for long time ignored in the study of hydrogen sulfide adsorption-oxidation. However, Kaliva and Smith [37] indicated that water likely forms complexes with oxygen thus taking part in the surface reaction, the effect of carbon surface chemistry, besides the influence of ash, was not discussed in details. Surface chemistry of carbons is rarely taken into account and it is even not considered as one of the required specifications for many applications. It is well known that the degree of acid dissociation depends on the pH of the system, and dissociation is feasible when pH is greater than pKa of an acid under study. Since hydrogen sulfide is a weak acid, analysis of the performance of carbons showed the dependence of the capacity on the acidity of carbon [38, 41, 48]. Moreover, threshold values were found on the dependence of the parameters describing the acidity of carbons and the normalized (for pore volumes) H 2 S breakthrough capacity values [48]. This clearly showed that the local pH in the pore system has a significant effect on the efficiency of hydrogen sulfide dissociation and thus its oxidation to various sulfur species. A moderately low average pH of the carbon surface is expected to suppress the dissociation of H 2 S and the formation of hydrogen sulfide ions. Those ions, when present at low concentration in small pores, are oxidized to sulfur oxides from which sulfuric acid is formed. On the other hand, a pH in the basic range promotes the dissociation ofH 2 S. This results in a high concentration ofHS- ions, which are then oxidized to sulfur radicals and polymers having chain or ring-like shapes. When the pH value is very low only physical adsorption can occur. The dependence of the normalized capacity on the pH of the carbon surface
is presented in Fig. 21.1 The threshold value derived from the analysis of the
,•
300 M
-
250
5
200
E () -.... 0)
•
•
•
~
"0
ca a. 150 ca ()
"0 Q)
"~ (ij
E 0
Z
100 50 0
0
2
4
6
8
10
12
Surface pH
Figure 21.1 Dependence of normalized H 2 S breakthrough capacity (per unit pore volume of carbon) on the surface pH (as described by Bandosz and coworkers [43, 48]).
21.2
Adsorption of Inorganic Gases
539
data occurs at the pH value around 4.5.The justification for the threshold in surface pH is based on the steps of hydrogen sulfide adsorption-oxidation on unmodified carbons [41, 48]. They are as follows: (1) H 2 S adsorption on the carbon surface, (2) its dissolution in a water film, (3) dissociation of H 2 S in an adsorbed state in the water film, and (4) surface reaction with adsorbed oxygen. The effect of ash can be considered as an extension of the effect of surface chemistry, since the pH of carbon expresses the average number and strength of acidic groups. While studying the hydrogen sulfide uptake on various carbons, it was found that presence of iron oxides or metals ions from group 6-8 has an effect on hydrogen sulfide adsorption [19, 49, 58, 59]. It is not only reflected in the amount adsorbed but also in the extent of oxidation. According to Steijns and Mars [19], the presence of iron oxide promotes formation ofS0 2 when removal process occurs at elevated temperatures. That effect was also noticed for removal of hydrogen sulfide on carbons at ambient conditions [32, 49]. The study of hydrogen sulfide removal on coal fly ash showed the noticeable adsorptionoxidation on those materials where active components for catalytic oxidation are mullite, hematite, and magnetite, all containing iron oxides [58, 59]. This effect was also clearly seen when carbonaceous materials derived from sewage sludge were tested as hydrogen sulfide adsorbents [64]. On them, an exceptionally high adsorption capacity, higher than that on coconut shell-based carbon was found. That superior performance was attributed to the catalytic reactions on ash, in particular on iron, copper and zinc oxides. When removal of hydrogen sulfide at 820 K was studied on carbons impregnated with zinc and copper a significant increase in the capacity was noticed; however, the activity diminished when the temperature of process was lowered [60]. Analysis of the above factors helped to formulate to the pH-dependent mechanism of hydrogen sulfide adsorption on activated carbons [41, 48]. It is summarized in Fig. 21.2 where the link exists between the extent of dissociation and the products of hydrogen sulfide adsorption. When the environment is moderately basic an increase in the concentration of HS- occurs. When the pH is distinctively acidic, the concentration of hydrogen sulfide ions is very low. In such situation hydrogen sulfide ions - when adsorbed in small pores are oxidized and converted to highly dispersed sulfur. These separated sulfur "islands" are susceptible to further oxidation to S02 and S03. This may happen since at such conditions the probability that an isolated sulfur atom will meet its own species is low. The main source of oxygen is likely oxygen from air. Its active radicals are adsorbed on the surface and accept electrons from sulfur. When the pH is less acidic (more basic) the concentration of HS- is much higher, which forces the created sulfur atoms to be close to each other, capable of forming polysulfides [28]. Then their polymerization to stable chain or cyclic sulfur molecules such as Ss occurs. It is apparent that further oxidation of such sulfur species is less probable than in the case of atomic sulfur. Summarizing, the results of the complex process of oxidation and its yield depend on the specific environment inside the porous structure of carbon. This environment consists of
Chapter
54°
21
Removal of Inorganic Gases and VOCs on Activated Carbons
S02(ads) + 0.5 O2 --+ S03(adS)
Cf free active sites Cf + 0.50 2 --+ C{O)
C(SSH) + 2HS- ~ C(S3SH) + H20
I Figure
21.2
pH-dependent mechanism of H 2 S adsorption-oxidation (as described by Adib
eta!' [43]).
the combination of pore sizes, degree of activated carbon hydrophilicity (related to surface functional groups), and local pH. The latter is the overall effect of carbon surface chemistry or applied chemical modifications. One can talk about the "local pH effect" only when the process occurs in pores where the existence of the water film is possible and surface groups can dissociate. The best conditions leading to oxidation of hydrogen sulfide to S4+ or S6+ exist when the concentration of HS- is just right (not too high, not too low) to be oxidized to highly dispersed sulfur. On the other hand, when the content of H 2 S0 4 rises in the course of the experiment" it suppresses the dissociation of hydrogen sulfide
21.2
Adsorption of Inorganic Gases
54 1
and inhibits the adsorption process. Under these conditions the small amount of H 2 S absorbed in the film of acid can be oxidized to elemental sulfur. The described above mechanism is also true when adsorption of hydrogen sulfide on nitrogen-containing carbons is discussed [50-57]. Such materials were introduced by Calgon Carbon in their proprietary process of Centaur® preparation [53]. Centaur®, a catalytic carbon that targets hydrogen sulfide removal is prepared by introducing the basic nitrogen functionality to the small pores of adsorbent. Although is done using impregnation with urea followed by high-temperature heat treatment, other nitrogen-containing organic compounds can be used [51, 54]. That treatment results in the presence of quaternary and pyridine-like nitrogen in the small pores [57]. As indicated above, basicity of such species, in the presence ofmoisture, enhances hydrogen sulfide dissociation, adsorption, oxidation, formation of radicals, and then their oxidation to sulfuric acid. However the total capacity of Centaur® is not exceptionally high (around 0.060 g/cm3 ) [57], its surface conversion of hydrogen sulfide to sulfuric acid is almost complete. This makes regeneration of spent materials using simple water washing feasible [50, 51]. Although nitrogen modification and oxidation to sulfuric acid makes Centaur® a superior product, high costs and risks related to the removal of concentrated acid from the surface limits its industrial and municipal applications in favor of caustic-impregnated or virgin activated carbons [9].
Table 21.1 H2 S breakthrough (ASTM 06646-01) capacities on few commercial activated carbons [65]
WVA 1100 (Westvaco) BAX 1500 Xtrusorb 60 (Calgon Carbon) Maxsorb (Kansai) 208c (Waterlink Barnabey and Sutcliffe) G55C (PICA) ROZ3 (Norit) Vapure 612 (Norit) R 2030(Norit) RB4 (Norit) STIX (Waterlink Barnabey and Sutcliffe) PCB-O (Calgon Carbon) BPL F3 (Calgon Carbon) MVP (Calgon Carbon) BPL 4xl0 (Calgon Carbon) FCA (Calgon Carbon) IVP (Calgon Carbon) Centaur HS® (Calgon Carbon) Centaur® (Calgon Carbon)
0.014 0.025 0.011 0.003 0.026 0.027 0.100 0.068 0.037 0.032 0.100 0.011 0.031 0.079 0.023 0.204 0.194 0.159 0.090
54 2
Chapter
21
Removal of Inorganic Gases and VOCs on Activated Carbons
Table 21.1 summarizes the HzS breakthrough capacity results obtained on few commercial activated carbons. The results are done using the same tests (ASTM D6646-01) so the comparison is meaningful. 21.2.2
Removal of Sulfur Dioxide
An increase in the acidity of natural water, fast rate of abrasion of buildings and monuments and health problems associated with this caused that desulfurization of fossil fuels and SOz removal from flue gases are technologies which have been developing rapidly during the last 20 years. Efficient media for removal sulfur dioxide are activated carbons [66-84] and activated carbon fibers [72,78,82]. Numerous studies indicate good efficiency of SOz removal on these materials either at low [72, 77, 78, 74, 76] or high temperatures [70,71, 80, 81]. Process ofSO z adsorption has been studied extensively and, like in the case of hydrogen sulfide, such parameters as porosity [68-77, 85, 86], surface chemistry [68, 71, 73, 75, 77, 78, 80, 82], and constituents of ash [81-89] were taken into consideration [5-16]. The products of surface reactions were analyzed from the point of view of removal efficiency and the feasibility of regeneration [74, 85]. Due to the higher oxidation state of sulfur in SOz than in HzS, the chemistry involved in immobilization is expected to be much less complex than that for oxidation of hydrogen sulfide. Since usually the process is carried out in the presence of moisture and oxygen, it is a generally accepted that sulfur dioxide is oxidized to sulfuric acid as a final product of the reaction. That acid is strongly retained in the pore system of activated carbons. Higher extent of oxidation usually results in more 50 z adsorbed [76]. Adsorption-oxidation of 50 z in oxygen atmosphere and in the presence of water occurs as follows [76]: SOZgas --+ SOZads
(21.8)
°Zgas --+ 20 ads
(21.9)
SOZads+Oads --+ S03ads
(21.10)
°
Hz gas --+ HZO ads S03ads + HZO ads --+ HZS04ads'
(21.11) (21.12)
where indices "gas" and "ads" refer to the presence of reactants in the gas phase and the adsorbed state, respectively. It was also found that three forms of adsorbed sulfur oxides could be present in such a situation. They are: weakly adsorbed SOz, physically adsorbed S03 (after oxidation of SOz), and strongly adsorbed H ZS0 4 [74-78]. It is well known that small pores, similar in size the adsorbate molecule, enhance the adsorption potential resulting in strong adsorption forces. Moreover, in the case of adsorption of sulfur dioxide it was demonstrated that oxidation to sulfur trioxide occurs mainly in the 7 A pores [76]. With an increase in the
21.2
Adsorption of Inorganic Gases
543
size of pores less 50 2 is converted, which results in smaller uptake of sulfur dioxide. No correlation was found between the amount of 502 adsorbed in the presence of oxygen and the volume of micropores. On the other hand, in the absence of oxygen the influence of pore volume was more pronounced. This can be explained by the extent of oxidation, which is enhanced when oxygen is present in the system. Supporting for this are the studies of the energetics of the adsorption where it was found that sulfur dioxide is adsorbed with two adsorption energies on activated carbons [68-77]. The low energy, about 50 kJ/mol, corresponds to weak physical adsorption, and the second, about 80 kJ/mol, to chemisorption [76], likely in the form of sulfuric acid. A significant effect of very small micropores on S02 adsorption was also noticed by Bagreev and coworkers [85]. The evidence on adsorption of sulfur dioxide in micropores in the absence of oxygen was found by Molina-Sabio and coworkers [77]. While calculating the micropore volumes of various carbons using CO 2, N 2, and S02' a relatively good agreement in the values was obtained. A small discrepancy found in the case of S02 was explained by its polarity. The strong adsorptive-adsorptive interaction in the gas phase caused weaker adsorbent-adsorbate interaction than in the case ofN2 and CO 2. Similar effect on micropore filling mechanism was also noticed by Wang and Kaneko [86]. Daley and coworkers [78] found that the S02 adsorption capacity on the activated carbon fibers was inversely proportional to pore size, pore volume, and pore size distribution. Although effects of porosity are crucial for physical adsorption, when weak adsorption forces exist, the importance ofthe catalytic effects ofsurface chemistry increases. In the case of acidic gases such as S02' the positive effect on adsorption should be observed when the basicity of surface increases. Numerous researches found that heat treatment of activated carbons or activated carbon fibers at temperatures about 1300 K results in an increase in the amount of sulfur dioxide adsorbed [68-74]. Such treatment, besides removal of oxygen-containing acidic groups, should increase carbon basicity [90]. It was specifically found that when basic groups containing oxygen are present on the carbon surface the adsorption ofS0 2 is significantly enhanced [73, 75]. In such a case basic groups (pyronic and pyronic-like type) are responsible for strong physical adsorption ofsulfur dioxide. Of course, those acid-base interactions do not introduce any catalytic effect leading to the formation of sulfuric acid and its chemisorption on the surface. The strong adsorption of sulfur dioxide is enhanced by the presence of oxygen [68, 71, 73, 75, 78]. These oxygen-containing sites are proposed to act as catalytic centers for oxidation ofS0 2 to S03 [79]. According to Davini [71, 73] oxygen present in the system plays an important role in the variations of S02 adsorbed. The negative role of oxygen in the amount of S02 adsorbed is linked to its ability to react with carbonaceous matrix, formation of surface groups, which decrease the surface area of adsorbent. A decrease in the S02 uptake upon the presence of oxygen-containing acidic groups was also noticed by Daley and coworkers [78]. They found the correlation between an increased S02 capacity and the amount of CO-C0 2 evolved during heat treatment of
544
Chapter
21
Removal of Inorganic Gases and VOCs on Activated Carbons
carbon fiber surfaces, which led to the formation of new active centers. Similar correlation was investigated by Mochida et al. [82]. On the other hand, Daley and coworkers [78] found that when dry S02 was adsorbed, the presence of oxygen-containing functional groups significantly enhanced the performance at temperature smaller than 348 K. That enhancement was explained by surface reactions of quinines with S02 and water forming diol and sulfuric acid. The effect of surface chemistry on S02 oxidation step was also discussed in details by Raymundo-Pinero and coworkers [76]. However, contrary to Daley and coworkers, they did not find any correlation between the amount of groups decomposed during heat treatment and an increase in the S02 adsorbed, they confirmed that removal of oxygen form the surface forms new high-energy adsorption-oxidation centers. An increase in the uptake of S02 upon oxidation of carbon was found by Lisovskii and coworkers [74]. It was postulated that surface acidic groups are the catalyst for S02 oxidation. Moreover, the presence of strong basic functionality was suggested as not beneficial for the process of sulfur dioxide removal due to an increase in the retention of sulfuric acid, which is undesirable form the point of few of adsorbent regeneration [74]. High adsorption on chars containing high surface acidity and basicity was also noticed by Rubio and Izquierdo [80]. Based on the performance of their materials, they concluded that not only the amount of surface groups but also their accessibility have an effect on S02 adsorption-oxidation. Besides empirical studies, ab initio molecular orbital calculations were performed on the possible pathways of S02 oxidation on activated carbon in the presence of water and oxygen. Yang and Yang [91] found that when surface oxides are present on the zigzag edge sites sulfuric acid is formed with sulfurous acid as an intermediate. On the other hand, chemisorption was found unfavorable on the edge sites containing twin oxides. Those findings can help to explain the discrepancies described above regarding the role of surface oxygen groups in 502 adsorption-oxidation on activated carbons. Basic nitrogen species present on the surface of activated carbons or carbon fibers, like in the case ofH25, were found to enhance the sulfur dioxide uptake. Polyacrylonitrile (PAN)-based activated carbon fibers are examples of good adsorbents for S02 removal [82, 92]. Although role of nitrogen present in the carbon matrix was not emphasized by Lee and coworkers [93] in their studies of S02 adsorption on PAN-based activated carbon fibers, [93] Kawabuchi and coworkers noticed a significant increase in the sorption capacity when activated carbon fibers were modified with pyridine and basic nitrogen functionalities were introduced to the surface [93]. Pyridine provided basic functionality, which increased catalytic removal of SO x. The effect ofintroducing nitrogen functionality to activated carbon surfaces on S02 removal was also studied in details by Raymundo-Pinero and coworkers [76] and Bagreev and coworkers [85]. Both group of researchers found that nitrogencontaining pyridinic species, which are placed at the edges of graphene layers, noticeably increase the amount of 50 2 adsorbed and its catalytic conversion to
21.2
Adsorption of Inorganic Gases
545
sulfuric acid. The effect is even more pronounced when those groups are present in small pores [85]. The only negative part related to the application of these materials is strong adsorption ofsulfuric acid leading to the difficulty in adsorbent regeneration, which was mentioned earlier by Lisovskii and coworkers [74]. As in the case of hydrogen sulfide, the presence of ash and its composition should have an effect on the amount of 50 2 retained on the surface. This effect was observed by Lu and Do [87] studying the 50 2 adsorption on activated coal rejected char. Its high content of inorganic matter-inorganic oxides was expected to affect the amount of 50 2 oxidized to sulfuric acid. As the most active ingredients, titanium oxide was identified. The enhancement in the oxidation of 502 due to the presence of active inorganic matter was also found by Bashkova et al. [89] on carbonaceous adsorbents derived from sewage sludge. In those materials, a high content of CaO was identified as a favorable factor. The effect of calcium was also studied when fly ash mixtures with calcium hydroxide were tested as 502 adsorbents [89]. It was found that Ca(OH)2 enhances the dispersion of calcium reagent and thus improves the efficiency of the adsorbent. 21.2.3 Adsorption of Hydrogen Cyanide
Another inorganic gas of extremely high toxicity is hydrogen cyanide (HCN). Besides its application as a warfare gas, called prussic acid (WWI) , or a massive extermination mean (WWII-cyclone B), it is an important industrial agent used in extraction of precious metals, in electroplating industry, metallurgy, and in the production of such materials as plastics, fire retardants, cosmetics, dyes, pains, and pharmaceuticals. As in the case of hydrogen sulfide and sulfur dioxide, physisorption of hydrogen cyanide on activated carbons is poor. Taking into account various chemical compounds, efforts were made to develop impregnated carbons, or carbon fibers (cloths) with efficient strength of the bonds between the irnpregnant(es) and the carbon surface [94-100]. On such materials chemisorption is the predominant process. Taking into account the toxicity of HCN, it is important that the removal process is irreversible and the adsorption capacity is sufficiently high. Usually efficient systems to remove HCN contain copper and an oxidizing salt such as sodium dichoromate [94]. Copper, either from nitrate or oxide is reduced to metal, then the oxidizing agent is added. Since the mixture is expected to exist in mesopores, they are active in the adsorption-chemisorption process. The chemistry of reactions in such systems was proposed by Alves and Clark [94]. It occurs in two stages: Cr+
Cr+
NC - CN + 2H 20 ~ CN - CONH 2 + H 20 ~ (CONH 2)2 cyanoformide oxamide (21.13)
They found that (CN)2 is an intermediate formed during HCN uptake on Cu-Cr-containing carbons. Then, in the presence of Cr6 + hydration occurs
546
Chapter 21 Removal of Inorganic Gases and VOCs on Activated Carbons
leading to oxamide, as the major product. On the other hand, on only copperimpregnated carbon the bulk of chemisorbed HCN is in the form of copper cyanide with no oxamide formation. In the case of Cu-Cr carbons the intermediate of the reaction, (CN)2 was found to be a factor enhancing the HCN removal capacity when compared to that on Cu carbon. Although copper, which is the most popular impregnant, results in a high HCN capacity (about 100 mg/g at 80 % humidity and 2 mg/L HCN in the challenge gas), the carbons containing the copper-chromate mixture are known to age with diminishing performance [95]. This directed the research efforts toward other metals-slats, which, when deposited on the carbon surface, can form complexes with HCN and cyanogens. The removal capacity on nickelor cobalt acetate-impregnated carbons was studied [95]. The uptake of90 mg/g was reached and the good breakthrough performance could be maintained for a year. The mechanism proposed was similar to that on Cu-impregnated carbons. It is believed that HCN dissolves in a saturated solution of metal acetate on the carbon surface and then metal-cyanide complexes and insoluble metal cyanides are formed [95]. Besides acetates, formates and propanoates of manganese, cobalt, nickel, copper, and zinc were used as impregnates ofactivated carbon cloth [96]. Although the exceptional capacities were not found, the results obtained supported the hypothesis that the reactions occurs in the film of water adsorbed on the surface and their pathways follow the reactivity of the impregnates with HCN in the solution. Determination of the participation of separate components of the systems in chemisorption and total adsorption was carried out by Rajakovic and coworkers [97]. They studied the performance of materials impregnated with compounds containing copper, silver (I), iron, magnesium, and aluminum in the forms of acetates, oxalates, tartrates, stearates, and citrates. Twofold structural activity of activated carbon cloth (ACC) was described. In spite of the fact that ACC is able to adsorb HCN without any impregnantes, adding metal salts acts chemically toward formation of complex compounds and precipitates. Although adsorptive and chemical forces are the most important factors, which affect the active bonding of pollutants, there are other important aspects influencing the performance of materials. They are the geometry of the system (porosity), diffusion rate, and the kinetics of surface reactions. Moreover, a significance of the type of metal salt was indicated. Application of organic salts with large organic moiety as impregnantes enhances the performance compared to the inorganic counterparts. This is the result of the affinity of the organic part to graphene layers and its strong retention on the surface. It was also found that at higher concentrations of pollutants formation of complex compounds predominates, whereas at lower concentration, the precipitates are formed.
21.2.4
Adsorption of NOx
The main source of nitric oxides (NO x) is combustion of fossil fuel where the concentration of NO x in the exhaust gases is usually smaller than 1000 ppm.
21.2
Adsorption of Inorganic Gases
547
First NO is formed and then it is oxidized in the atmosphere to N0 2 • Since in combustion, the origin of nitrogen is not only from N-rich fuel but also from air supplied for oxidation, in the elimination of NO x postcombustion methods are important. So far the most effective technique has been selective catalytic reduction of NO x on various catalysts. When the activated carbons are used as removal media, the elimination process includes also adsorption combined either with oxidation or reduction. Oxidation usually leads to the formation of nitric acid whereas N 2 is the product of NO x reduction. As in the cases of other pollutants addressed in this review, for NO x removal either unmodified or impregnated (caustics, catalytic metals) activated carbons have been used [101-114]. Adsorption of NO x on unimpregnated activated carbons has been studied at temperatures between 295 and 400 K and the pressure ranges between 100 and 3000kPa [101-105]. The reported amount of NO x adsorbed reached 150mg N0 2 per gram of carbon [101]. It was found that an increase in the pressure significantly increased the N0 2 uptake and that effect compensated the negative effect of an increasing temperature. As the mechanism of adsorption, the micropore filling was proposed with a significant amount of N0 2 adsorbed irreversibly. Nitric oxide was formed on the surface of carbon where catalytic reaction of oxidation of NO to N0 2 in the presence of oxygen was suggested to take place [101]. Adsorption of NO and its reduction on chars was studied by Izquierdo and Rubio [102]. The removal process was designed to work at the temperature range 373-473 K in the presence of oxygen and moisture. The removal capacity was between 100 and 150mg/g with around 30% of NO conversion. The conversion of NO via direct reduction with the carbon surface occurred as follows [102]: 2NO + C ---* N 2 + CO 2
2NO + O 2 + 2C ---* N 2 + 2C0 2
(21.14) (21.15)
It was found that surface chemistry affects the NO removal performance and an optimal amount of oxygen functional groups on the surface of char is needed. This amount has to be established experimentally. Izquierdo and Rubio [102] proposed that gas phase oxygen reacts with the carbon surface forming oxygen-carbon structures, which act as active centers for NO chemisorption. When the reduction proceeds, CO 2 is released and the new oxygen groups are formed. In the absence of oxygen, the conversion of NO decreases to zero when all active centers, functional groups are consumed. The effect of thermal surface treatment of carbons carried out in various atmospheres on the NO x adsorption-reduction was studied by Xia and coworkers [103]. They found that activated carbon heated at 1200K in hydrogen adsorbed NO without its oxidation to NO x • This happens as a result of a decreased affinity of the hydrogen reduced carbon to chemisorbed oxygen. It is interesting that such effect was not found on carbons treated in nitrogen.
548
Chapter
21
Removal of Inorganic Gases and VOCs on Activated Carbons
The amount of NO adsorbed on the hydrogen reduced carbons was about 1 mmoll g and the existence of strong (> 40 kcallmol) and weak chemisorption with various energies of adsorption was found. An enhancement in the process of NO x abatement with carbon adsorbents was proposed by Kong and Cha [104, 105]. Following the physical adsorption of NO x on unimpregnated carbons, treatment with microwave energy of 480 W was proposed. This process enhanced NO x reaction with carbon to produce nitrogen and carbon oxides. The few runs of microwave regeneration increased 10 times the surface area of spent char as a result of the activation process. Moreover, the nitrogen compounds were introduced to the carbon matrix changing its surface chemistry. Those stable compounds formed during dissociative chemisorption of NO were indicated to decompose to N 2 under microwave radiation. Kong and Cha concluded that in the presence of water and oxygen NO is converted to N0 2 and HN0 3 • During microwave treatment those species are reduced back to NO and the reaction with carbon occurs with formation of N 2 , CO, and CO 2 , A significant advantage of this process is that 90 % of NO x is reduced to N 2 and microwave treatment-regeneration can in fact be considered an activation method for low surface area chars. The effects of impregnation of activated carbons with potassium hydroxide on the efficiency of NO x removal was studied by Lee and coworkers [106, 107]. They found that KOH creates the selective adsorption sites (increases basicity of carbons by the presence of OH-) for NO x adsorption. As a result of the surface reaction, KN0 2 and KN0 3 are formed. Formation of salt crystals blocks the porosity of the materials and diminishes the NO x removal capacity. It was proposed that the surface basic OH- ions delay oxidation of KN0 2 to KN0 3 and thus result in an increase in the surface adsorptivity. The effect of potassium in the form of potassium carbonate or potassium silicate on reduction of NO x on coal chars was also investigated [108-111]. The best materials were prepared by pyrolysis of coal at 1300 K with high KOHl coal ratio [108]. On these adsorbents, at temperature smaller than 473 K, physical adsorption is predominant while the true NO x reduction by char occurs at T> 473K with formation ofN2 and CO 2 , The results indicated that a material with the high surface area should be used to promote adsorption of NO x and potassium remaining in chars catalyzes NO x reduction in the presence of oxygen [109]. The reduction of NO x on carbons can be also enhanced by the presence of ammonia [102, 112] or nitrogen-containing groups on the surface of carbons [104, 105, 113]. Mochida and coworkers [112] found that the presence of ammonia adsorbed on the surface of carbon fibers enhances the reduction of NO; however, the process is not efficient at humidity higher than 60% [113]. When ammonia is introduced to the reaction and oxygen is present the following reactions occur [102]: 6NO + 4NH3 -+ 5N2 + 6H 2 0
(21.16)
4NO + 4NH 3 + O 2 -+ 4N2 + 6H 2 0
(21.17)
21.3
Adsorption of Volatile Organic Compounds
549
Matzner and Boehm [113] found that the incorporation of nitrogen in activated carbons enhances their reduction activity toward nitric oxides. On such materials, reduction occurs at much lower temperature than that on undoped carbons and the conversion is higher. Moreover, the amount of NO adsorbed on nitrogen-doped carbon at room temperature increased with an increase in the content of nitrogen. As suggested, NO reacts with surface sites of chemisorbed nitrogen, C(N) and oxygen-containing site and nitrogen are formed: C(N) + NO -+ C(O) + N 2
(21.18)
It was proposed [113] that chemisorption reaction may also be associated with an electron transfer from the carbon surface to the NO molecule or (NO)2 molecule. The resulting species are diamagnetic and dimeric. Those hyponitrites are highly reactive and they can easy oxidize the carbon surface resulting in formation ofN2. Besides caustic and nitrogen modifications of the carbon surfaces, the reduction of NO x was extensively studied on carbons impregnated with transition metals [114-118] such as Ni, Fe, Co, or Cu. From all of those metals, copper was found as the most efficient catalysts toward reduction of NO into N 2 and O 2 either with or without oxygen. On the carbon - copper catalysts at temperature over 600 K 100 % conversion is reached with a high capacity of the adsorbent. It was found that a metallic catalytic system of NO-Cu reaction is very predominant and copper metal is activated for removing NO at high temperature even in the absence of oxygen [114]. As indicated above, in the studies of unmodified carbons, oxygen from NO-C reduction creates more active sites for NO adsorption via formation of high surface energy active sites on the surface.
21.3 ADSORPTION OF VOLATILE ORGANIC COMPOUNDS Volatile organic compounds commonly known as VOC are a group of various small molecule organic species with low boiling [119]. Among numerous organic species that can be considered as VOCs, US EPA lists 188 volatile organic compounds as the dangerous air and water pollutants. In these groups, one can find a broad spectrum of organic compounds from chlorinated species through ketones, aldehydes, carboxylic acids, and thiocompounds. Removal of those compounds occurs usually using condensation, absorption, oxidation, incineration, and adsorption. The last method with activated carbons as adsorbents is widely applied in industrial processes. This is owing to the predominantly hydrophobic nature of VOCs granting their interactions with activated carbon surfaces. Another important factor is a small size and high volume of activated carbon pores. They result in strong adsorption forces, even if traces of VOCs are present.
55°
Chapter
21
Removal of Inorganic Gases and VOCs on Activated Carbons
To study the adsorption of all listed VOCs on activated carbons is an overwhelming task. The experimental difficulties are linked to the long list of species to study and their high level of toxicity. This causes that the results described in the literature usually focus on one or few VOCs compounds as models, and those species are not necessary listed as the most dangerous pollutants by EPA. The tasks of research are usually established knowing the chemistry of the molecule to be studied and taking into account the existence of dispersive and specific interactions, which can playa role in the adsorption. Another important factor, which has to be taken into consideration is the concentration range, phase from which adsorption is supposed to occur and the temperature of the process. So far it has been determined that the equilibrium adsorption isotherms of such species as toluene, 1- butanol, and ethyl acetate can be described by the Langmuir-Freundlich or Toth equation [120]. The classical volatile organic compound whose adsorption was studied in details on activated carbon is benzene. In fact, before strict environmental regulations, benzene was used as a model compound to determine the porosity of activated carbons [121] owing to its stability, hydrophobicity, and symmetrical size of the molecule. The characteristic energy of adsorption used as a reference is calculated using the physiochemical properties of the benzene molecule. The interactions of various organic compounds, now considered as VOCs, with the surface of various activated carbons were also studied in details by Kiselev and Yashin [122]. Using inverse gas chromatography (IGC), they determined the energetic parameters (energy, enthalpy, entropy) of their interactions depending on the size of the molecule, sterical hindrances, and the heteroatoms present. The results obtained by them are very extensive and certainly can be used as a reference when adsorption of any VOCs is studied. The research of Kiselev and Yashin was done either on various nonporous carbon blacks or on porous activated carbons. Taking into account that the adsorption energy should double in micropores compared to the flat carbon surface, the accessibility of pores of the adsorbents for the molecules to be removed and thus the efficiency of the adsorption process can be estimated using these data. Since the fundamental studies of the Russian schools of Dubinin and Kiselev were carried out in the 1960s, the change in the approach of science toward application and concerns about environmental pollutions, directed the focus of scientific research toward studies of the adsorption of VOCs on carbons from the point of view of the feasibility of the removal process. Of course, this feasibility is related to the strength of the interactions, especially important at low concentrations, and the adsorption capacity (for high concentrations). Since that time also more has been done to understand specific interactions with the carbon surfaces, especially those decorated with functional groups containing heteroatoms [62, 63]. Recently, the effectiveness of removal of bromo-dichloromethane, benzene, carbontetrachloride, 1,1, i-trichloromethane, chloroform, and 1,1dichloromethane was studied on various laboratory-based and commercial activated carbons [123]. The pecan shell- and almond shell-based materials obtained
21.3
Adsorption of Volatile Organic Compounds
55 1
by either physical (steam, CO 2) or chemical (phosphoric acid) activation were used. For comparison, coconut shell and bituminous coal-based carbons were also studied. The obtained results showed, as would be expected, the superior adsorption of benzene compared to other species. The best performance for other VOCs studied was obtained for coconut and pecan shell physically activated carbons, which can be linked to their small pore sizes. All other VOC studied are halogenated compounds and they should interact with carbons only in a dispersive way. The removal of a broad range of environmentally detrimental VOCs on activated carbons was studied by Le Cloirec and coworkers [124]. Their extensive study led to quantitative relationship, which can be used to predict the energetic interactions resulting from either adsorption or desorption of VOCs on granulated activated carbons. To obtain that relationship, the adsorption of 40 VOCs was investigated using differential scanning calorimetry coupled to thermogravimetry. Multiple linear regressions were applied to correlate the data obtained to the physicochemical properties of the molecules. It was found that ionization potential, polarizability, and connectivity index have main influence on the adsorption energy of those species. Taking into account the difficulties in measuring the adsorption isotherms of VOCs, the obtained relationship, however simplified, can prove to be very useful in environmental engineering applications. The effects of physicochemical properties of VOCs on their adsorption capacity on activated carbons were also noticed by Chiang and coworkers [125]. Such properties as their boiling point, critical temperature, cross-sectional area, and dipole moment were found as the most important features governing activated carbon adsorption. Le Cloirec and coworkers [126] also studied the warming of the activated carbons adsorbent bed occurring during the removal process at high concentrations. The exothermal nature of the adsorption quite often results in bed ignition. It happens especially when very microporolls carbon is used, as for example coconut shell-based, or such species as ketones, aldehydes, carboxylic acids, or sulfur-containing species are to be removed. The heat effects in the cases of those species are not only related to the enhancement of adsorption potential due to physical adsorption but also are related to chemical reactions such as oxidation or dimerization taking place on the carbon surfaces at the temperatures even close to ambient [127]. To avoid self-ignition of the bed, the rise in temperature caused by oxidation of a solvent or carbon bed should be estimated and based on this, the proper maximum concentration of VOCs should be determined. It was also found that that at high VOCs concentrations, the moisture content of the air does not affect the carbon capacity for VOC removal or warming of the bed [126]. The reason for this lies likely in the difference in the affinity of water and VOC to be adsorbed on carbons. The latter, having a hydrophobic moiety interacts with graphene layers much stronger than water. Water, even if preadsorbed, is likely replaced with the organic molecule when the removal process proceeds. On the other hand, Chou and Chiou [128] in their research of the removal ofVOCs from exhausted gas stream found that
552
Chapter
21
Removal of Inorganic Gases and VOCs on Activated Carbons
moisture content in gas is unfavorable for the removal process. Similar effect was observed by Shin and coworkers [129] in their study of adsorption of benzene, toluene, and ethyl benzene. When relative humidity reached 60 % the adsorption of those species on activated carbon significantly decreased. For the multicomponent adsorption of VOC, displacement effects were noticed and adsorbates with strong interaction forced to displace weakly bounded species. The amount adsorbed was found to depend on the size of molecule (larger molecules - greater adsorption) and its polarity (less polar - greater adsorption). A detailed study of adsorption of aldehydes on activated carbons was performed by Domingo-Garcia and coworkers [130] and EI-Sayed and Bandosz [131, 132]. Using IGC at infinite dilution, it was found that formaldehyde is strongly adsorbed on activated carbons with isosteric heat between 15 and 33 kJ/mol [130] and the retention volumes increased with an increase in the surface areas of activated carbons. The results of the studies of adsorption of acetaldehyde [131] showed that the amount adsorbed depends strongly on the pore size distributions of carbons and their surface chemistry. When very small pores, close in size to the acetaldehyde molecule, and oxygencontaining groups are present (to certain extent), the heat of adsorption reaches its maximum value. A small density of surface groups can enhance the heat of adsorption whereas extensive oxidation leads to a decrease in the strength of adsorption forces. This happens due to the blocking of the pore entrances with functional groups and a decrease in the accessibility of the hydrophobic surface where the dispersive interactions of the hydrocarbon moiety with small pore walls can be enhanced. Oxidation of the carbon surfaces results in an increase in the amount of acetaldehyde adsorbed at saturation conditions indicating the importance of hydrogen bonding of adsorbate molecule with functional groups present on the activated carbon surface. Similar effect was noticed when the nitrogen enriched carbons were used as adsorbents [132]. Moreover, the adsorption capacity was found to depend strongly on the volume of pores in the adsorbents. The strong effect of dependence on the pore volume was also observed by Fuertes and coworkers [133]. For adsorption of n-butane and nhexane at room temperature it was found that at high adsorbate concentrations the amount adsorbed is a function of the pore volume, while at low concentrations, it depends mainly on pore size distributions of carbons. Moreover, at low relative pressure (P/Po < 0.004) the amount adsorbed can be correlated with the molecular parachor and the polarizability of adsorbates. The effects of surface properties of carbons on adsorption of amines [134, 135] and carboxylic acid were also investigated [136, 137]. For amines, at small concentrations, the acidic groups increased the amount adsorbed [135] whereas, in the case of valeric acid, the surface basic groups interacted with the adsorbate molecule [136, 137]. For the gross adsorption capacity, the volume ofmicropores, especially those smaller than loA governed the performance of materials. Study ofadsorption ofpolar methyl tertiary-butyl ether (MTBE) and nonpolar 1- methylbutane vapors on activated carbons in the dynamic conditions was carried out by Gironi and coworkers [138]. The maximum capacities ofactivated
21.4
Choice of Proper Carbon for a Desired Application
553
carbons for adsorption of air and MTBE or 1-methylbutane were equal to 0.55 and 0.45 gig, respectively. It was observed that during the adsorption of mixtures, MTBE is adsorbed preferentially on the carbon and the progressive saturation of the solid bed by MTBE causes the displacement of the previously adsorbed 1-methylbutane. Detailed investigation of interactions of MTBE and TCE (trichloroethane) with the surface of various activated carbon fibers was performed from aqueous phase by Li and coworkers [139]. Following differences in the sizes of the molecules, it was found that 7-10 A pores are preferable for TCE adsorption whereas MTBE is adsorbed primarily in 8-11 A pores. The authors concluded that the effective adsorbents to remove TCE and MTBE should be microporous with the pore sizes of about 1.3-1.8 times larger than the kinetic diameters of the target molecules. Moreover, the surface should be hydrophobic and the amount of heteroatoms such as oxygen and nitrogen should not exceed 3 mmol/g. The presence of surface functional groups is not favorable for adsorption of both those species due to the competition with water for high-energy sites. When adsorption was carried out from cyclohexane, surface of oxidized carbon was found to be preferable for MTBE adsorption. It was due to the preferential adsorption of MTBE on carboxylic acids and phenolic hydroxyl groups where hydrogen bonds could be formed between ether oxygen and hydrogen atoms of those groups. In water the adsorbents always exhibit a larger adsorptive capacity for TCE than for MTBE due to the greater aqueous solubility of MTBE. The importance of hydrogen bonding was also underlined in the studies of such volatile organic compounds as alcohols [140] or diethyl ether [141]. The strength of adsorption of those species increased when heteroatoms were incorporated to the matrix but for the gross adsorption capacity the volume of micropores was important. The chemistry of adsorbed VOCs molecules has also an effect on adsorption energy, which may increase with surface coverage when functional groups are present [142] due to the adsorbate-adsorbate interactions. Moreover, in some cases the chemical reactions of specific organic compounds with surface groups can occur in the presence of hot air as indicated by Popescu and coworkers [143].
21.4 CHOICE OF PROPER CARBON FOR A DESIRED ApPLICATION
The variety of activated carbons, carbon fibers, and carbon monoliths present on the market along with differences in the molecules to be adsorbedremoved causes that the choice of the adsorbents for a desired application becomes a difficult task. The capacities, for H 25, 502' NO x , HCN, or VOCs removal depend on the type of carbon used (Fig. 21.3). The problem is even more complex when multicomponent adsorption is expected to occur and the regeneration options have to be considered. Usually carbon specifications
Chapter
554
21
Removal of Inorganic Gases and VOCs on Activated Carbons
0.5 0.45 0.4
C> -...
.9
0.35
"C Q)
.0
0.3
"C ctS
0.25
(; en
C ~
0
E
«
0.2 0.15 0.1 0.05 0
Figure 21.3 Summary of the activated carbons-activated carbon fibers capacities for various species reported in the literature.
list such features as surface area (iodine number), density, hardness, and ash content, along with the specific test checking the target performance such as the hydrogen sulfide capacity or butane working capacity [13]. Those numbers, even if obtained following exactly the test procedures, should be compared with great precautions. For instance, in the case of hydrogen sulfide, to obtain meaningful results the experiments should be done at very low concentrations of H 2 S, taking into account the possibilities of the large differences in the rates oxidation on unmodified and impregnated carbons. Oxidation ofH 2 S on caustic carbon is a fast reaction while oxidation on unmodified carbon is rate limited due to the discussed above the complexity of the process. When the concentration is low and the contact times in the bed are long enough, both processes can go to completion. This means that the accelerated test, which is a standard procedure to evaluate the H 2 S breakthrough capacity of carbons, can be used for comparison of results only when the mechanisms of reactions are more or less similar. It follows that results can only be compared within the categories of unmodified carbons or caustic-impregnated ones. Moreover, it should be always taken into account that the conditions oflaboratory tests are different from those in real life. In the real environment, for instance, in sewage treatment plants, carbons are exposed to other species besides H 2 S, including many hydrocarbons, VOCs, and CO 2 • These species can enhance the breakthrough capacity by changing the pH of the carbon surface but they can also decrease the capacity by blocking the high-energy adsorption centers, small pores.
N ~
~
Table 21.2 Summary of important surface features governing the adsorption of inorganic species and VOCs at the temperatures close to ambient in the presence of air and humidity
()
::::r
or::;. (t)
S, ""'C
a
"'0
~
() Q)
H2S
S02 NO x
Physical adsorptiondissolution-oxidation
Physical adsorption-oxidation Adsorptionoxidation-reduction
HCN
Adsorptioncomplexationoxidation
VOC
Physical adsorption
Sulfur radicals, sulfur polymers, S02, H 2S0 4
S02, H 2S0 4 NO, N0 2, N 2, CO 2, CO, HN0 3 (CNh, CNCONH 2, (CONH 2)2 metal cyanides No reaction
Crucial component, adsorption in film of water, ensures dissociation
Volume of micropores, small size « 1 nm) promotes fonnation of S02 and H 2S0 4
Enhanced removal Should be less than 60%
Pore width 4.5, basic nitrogen groups significantly increase basicity Active centers-oxygen- and nitrogen-containing groups Oxygen- or nitrogen-containing groups are centers for NO chemisorption
Fe 2 0 3 , CaO Cu, metals from group
6-8
co ::::s
~
Q)
o(t) ~.
roc.. »
"'0
""£. r::;.
CaO,
~
o·
::::s
Cu
Cu, Ag, Fe, Ni, Zn Oxygen and nitrogen groups important for adsorption of small concentration of polar VOCs
V1 V1 V1
Chapter
556
21
Removal of Inorganic Gases and VOCs on Activated Carbons
As described above, very important factors influencing the performance of carbons as adsorbents of either inorganic gases or VOCs are their surface area, pore volume, and pore size distributions. They are important not only as active centers for physical adsorption but also as storage space for chemically enhanced adsorption (oxidation or complexation) as happens in the case of hydrogen sulfide, sulfur dioxide, or hydrogen cyanide. The specific stress should be put here on the pore size distribution since the pore sizes are really critical for removal of pollutants at low concentrations. The sizes should be comparable to the size of the molecule to be adsorbed (one to twice larger) to impose the strong adsorption forces. Besides porosity, it is clearly demonstrated the surface chemistry of carbons is also a very important factor influencing the adsorption and it should not be neglected. Even subtle changes in surface acidity and basicity can have an effect on the amount adsorbed or on the extent of chemical reactions occurring on the surface. Moreover, the ash content or the content of the catalytic metals in ash has been also indicated as affecting the performance of adsorbents. In some cases, to reach the final goal of surface reaction as, for instance, reduction of NO x by carbon, NO has to be first adsorbed and oxidized, and for this step the surface oxygen groups are very beneficial. In the case of VOCs, the situation is even more complex. They are organic compounds so even though their affinity to be adsorbed on the hydrophobic surface of carbon is an unquestionable fact, the surface chemistry plays a critical role when concentrations are very small and the molecules exhibit various degrees of polarity as a result of the presence of heteroatoms such as oxygen, nitrogen, or chlorine. At such conditions, the strength of specific interactions via hydrogen bonding or an acid-base mechanism can be important to enhance the amount adsorbed. Table 21.2 summarizes the specific features of carbon surface, which have been demonstrated as important for the adsorption processes of species addressed in this chapter. These data can be used as a specific guideline to find the best adsorbents for the desired applications. As mentioned above, when the adsorbed phase contains more then one component, which occurs in the majority of reallife applications, the possibility of physical and chemical interactions between various molecules and the effect of those interactions on the enhancementinhibition and the feasibility of the removal process should be considered.
REFERENCES 1. Manahan, S.E. (1997). Environmental Chemistry, 7th edn. CRC Press. 2. http:/www.epa.gov/air/oaqps/peg-caa/pegcaa05.html 3. http:/www.epa.gov/air/caa
References
557
4. ]irsak, T., Dvorak, ]., and Rodriguez, ].A. (1999). Chemistry of thiophene on Zn, S/ZnO, and Cs/ZnO surfaces. Effects of cesium on desulfurization processes. J. Phys. Chem., 103, 5550-6. 5. Turk, A., Sakalis, S., Lessuck, J., et al. (1989). Ammonia injection enhances capacity of activated carbon for hydrogen sulfide and methyl mercaptan. Environ. Sci. Technol., 33, 1242-5. 6. Stuetz, R.M., Fenner, R.M., and Engin, G. (1999). Assessment of odours from sewage treatment works by an electronic nose, H 2 S analysis and olfactometry. Water Res., 33, 453-61. 7. Turk, A., Mahmood, K., and Mozaffari, J. (1993). Activated carbon for air purification in New York City's sewage treatment plants. Water. Sci. Techno I. , 27, 121-6. 8. Turk, A. and Bandosz, TJ. (2000). Adsorption systems for odour treatment. In Odours in Wastewater Treatment: Measurement, Modeling and Control. (R.M. Stuetz and F.-B. Frechen, eds). IWA, pp. 354-64. 9. Bandosz, TJ., Bagreev, A., Adib, F., and Turk, A. (2000). Unmodified versus caustics-impregnated carbons for control of hydrogen sulfide emissions from sewage treatment plants. Environ. Sci. Technol., 34, 1069-74. 10. Yan, R., Liang, D.T., Tsen, L., and Tay, J.H. (2002). Kinetics and mechanisms of H 2 S adsorption by alkaline activated carbon. Environ. Sci. Technol., 36, 4460-6. 11. Chiang, H.-L., Tsai, J.-H., Tsai, C.-L., and Hsu, Y.-C. (2000). Adsorption characteristics of alkaline activated carbon exemplified by water vapor, H 2 S and CH 3 SH gas. Sep. Sci. Technol., 35,903-18. 12. Bagreev, A. and Bandosz, TJ. (2002). A role ofsodium hydroxide in the process of hydrogen sulfide adsorption/oxidation on caustic-impregnated activated carbons. Ind. Eng. Chem. Res., 41, 672-9. 13. ASTM Standards (1998). Vo1.15.01. Refractories; Carbon and Graphite Products; Activated Carbon; Advanced Ceramics. ASTM D6646-01. 14. Bandosz, TJ. and Le, Q. (1998). Evaluation of surface properties of exhausted carbons used as H 2 S adsorbents in sewage treatment plants. Carbon, 36, 39-44. 15. Przepiorski, J. and Oya, A. (1998). K2 C0 3 -loaded deodorizing activated carbon fibre against H 2 S gas: factors influencing the deodorizing efficiency and the regeneration method.]. Mater. Sci. Lett., 17,679-82. 16. Przepiorski, J., Yoshida, S., and Oya, A. (1999). Structure of K2 C0 3 -loaded activated carbon fiber and its deodorization ability against H 2 S gas. Carbon, 37, 1881-90. 17. Przepiorski, J., Abe, Y., Yoshida, S., and Oya, A. (1997). Preferential supporting of potassium carbonate around the peripheral region of activated carbon fibre. ]. Mater. Sci. Lett., 16, 1312-14. 18. Coskun, I. and Tollefson, E.L. (1986). Oxidation oflow concentrations of hydrogen sulfide over activated carbons. Can. J. Chem. Eng., 58, 72-6. 19. Steijns, M. and Mars, P. (1977). Catalytic oxidation of hydrogen sulfide. Influence of pore structure and chemical composition of various porous substances. Ind. Eng. Chem. Prod. Res. Dev., 16, 35-41. 20. Ghosh, T.K. andTollefson, E.L. (1986). A continuous process for recovery of sulfur from natural gas containing low concentrations of hydrogen sulfide. Can.]. Chem. Eng., 64, 960-8.
558
Chapter
21
Removal of Inorganic Gases and VOCs on Activated Carbons
21. Ghosh, T.K. and Tollefson, E.L. (1986). Kinetics and reaction mechanism of hydrogen sulfide oxidation over activated carbon in the temperature range of 125-200°C. Can. J. Chern. Eng., 64, 969-76. 22. Dalai, A.K., Majumadar, M., Chowdhury, A., and Tollefson, E.L. (1993). The effects of pressure and temperature on the catalytic oxidation of hydrogen sulfide in natural gas and regeneration of the catalysts to recover the sulfur produced. Can. J. Chern. Eng., 71, 75-82. 23. Yang, A., Tollefson, E.L., and Dalai, A.K. (1998). Oxidation of low concentrations of hydrogen sulphide: process optimization and kinetics studies. Can. J. Chern. Eng., 76, 76-86. 24. Dalai, A.K. and Tollefson, E.L. (1986). Kinetics and reaction mechanism of catalytic oxidation of low concentrations of hydrogen sulfide in natural gas over activated carbon. Can. J. Chern. Eng., 76, 902-14. 25. Dalai, A.K., Majumdar, A., and Tollefson, E.L. (1999). Low temperature catalytic oxidation of hydrogen sulfide in sour produced wastewater using activated carbon catalysts. Environ. Sci. Technol., 33, 2241-6. 26. Meeyoo, V., Trimm, D.L., and Cant, N.W. (1997). Adsorption-reaction Processes for the removal of hydrogen sulphide from gas streams.J. Chern. Tech. Biotechnol., 68,411-16. 27. Steijns, M., Derks, F., Verloop, A., and Mars, P. (1976). The mechanism of the catalytic oxidation of hydrogen sulfide. II Kinetics and mechanism of hydrogen sulfide oxidation catalyzed by sulfur. J. Catal., 42, 87-95. 28. Steijns, M., Koopman, P., Nieuwenhuijse, B., and Mars, P. (1976). The mechanism of the catalytic oxidation of hydrogen sulfide. III. An electron spin resonance study of the sulfur catalyzed oxidation of hydrogen sulfide. J. Catal., 42, 96-106. 29. Steijns, M. and Mars, P. (1974). The role of sulfur trapped in micropores in the catalytic partial oxidation of hydrogen sulfide with oxygen.]. Catal., 35, 11-17. 30. Klein, J. and Henning, K-D. (1984). Catalytic oxidation of hydrogen sulphide on activated carbons. Fuel, 63, 1064-7. 31. Hedden, K., Humber, L., and Rao, B.R. (1976). Adsorptive Reinigung von schwefelwasserstoffhaltigen abgasen. VDI-Bericht Nr. 253 S. 37/42, VDI-Verlag. 32. Katoh, H., Kuniyoshi, I., Hirai, M., and Shoda, M. (1995). Studies of the oxidation mechanism of sulphur-containing gases on wet activated carbon fibre. Appl. Catal. B Environ., 6, 255-62. 33. Tanada, S., Kita, T., Boki, K., and Kozaki, Y. (1985). Preparation of narrow pores carbon suitable for hydrogen sulfide adsorption. J. Environ. Sci. Health, A20, 87-96. 34. Choi, J.J., Hirai, M., and Shoda, M. (1991). Catalytic oxidation of hydrogen sulphide by air over an activated carbon fibre. App. Catal. A, 79, 241-8. 35. Sreeramamurthy, R. and Menon, P.G. (1975). Oxidation ofH 2 S on active carbon catalysts. J. Catalysis, 37, 287-96. 36. Primavera, A., Trovarelli, A., Andreussi, P., and Dolcetti, G. (1998). The effect of water in the low-temperature catalytic oxidation of hydrogen sulfide to sulfur over activated carbon. Appl. Catal. A Gen., 173, 185-92. 37. Kaliva, A.N. and Smith,J.W. (1983). Oxidation of low concentrations of hydrogen sulfide by air on a fixed activated carbon bed. Can.]. Chern. Eng., 61, 208-12. 38. Le Lauch, L.M., Subrenat, A., and Le Cloirec, P. (2003). Hydrogen sulfide adsorption and oxidation onto activated carbon cloth: applications to odorous gaseous emission treatments. Langrnuir, 19, 10869-77.
References
559
39. Mikhalovsky, S.V., Zaitsev, and Yu. P. (1997). Catalytic properties of activated carbons I. Gas-phase oxidation of hydrogen sulphide. Carbon, 35, 1367-74. 40. Bandosz, TJ. (1999). Effect of pore structure and surface chemistry of virgin activated carbon on removal of hydrogen sulfide. Carbon, 37, 483-91. 41. Adib, F., Bagreev, A., and Bandosz, TJ. (2000). Analysis of the relationship between H 2S removal capacity and surface properties of unmodified activated carbons. Environ. Sci. Technol., 34, 686-92. 42. Adib, F., Bagreev, A., and Bandosz, TJ. (1999). Effect of surface characteristics of wood based activated carbons on removal of hydrogen sulfide J. Colloid Interface Sci., 214, 407-15. 43. Adib, F., Bagreev, A., and Bandosz, TJ. (1999). Effect ofpH and surface chemistry on the mechanism of H 2S removal by activated carbons. J. Colloid Interface Sci., 216,360-9. 44. Adib, F., Bagreev, A., and Bandosz, T.J. (2000). On the possibility of regeneration of unimpregnated activated carbons used as hydrogen sulfide adsorbents. Ind. Eng. Chern. Res., 39, 2439-46. 45. Bagreev, A., Rahman, H., and Bandosz, T.J. (2000). Wood-based activated carbons as adsorbents of hydrogen sulfide: a study of adsorption and water regeneration process Ind. Eng. Chern. Res., 39, 3849-55. 46. Bagreev, A., Rahman, H., and Bandosz, TJ. (2000). Study of H 2S adsorption and water regeneration of coconut-based activated carbon. Environ. Sci. Technol., 34, 4587-92. 47. Bagreev, A., Adib, F., and Bandosz, T.J. (1999). Initial heats ofH 2S adsorption on activated carbons: effect of surface features. J. Colloid Interface Sci., 219, 327-32. 48. Bagreev, A., Adib, F., and Bandosz, T.J. (2001). pH of the activated carbon surface as an indication for its suitability for removal of hydrogen sulfide from wet air streams. Carbon, 39, 1987-905. 49. Bagreev. A. and Bandosz, T.J. (2001). H 2S adsorption/oxidation on unmodified activated carbons: importance of prehumidification. Carbon, 39, 2303-11. 50. Hayden, R.A. (1995). Process for making catalytic carbon. US Patent 5,444,031. 51. Bagreev. A., Menendez, J.A., Dukhno, I., et al. (2004). Bituminous coal-based activated carbons modified with nitrogen as adsorbents of hydrogen sulfide. Carbon, 42, 469-76. 52. Hayden, R.A. (1995). Process for regenerating nitrogen-treated carbonaceous chars used for hydrogen sulfide removal. International Patent WO 95/26230. 53. Matviya, T.M. and Hayden, R.A. (1994). Catalytic carbon. US Patent 5,356,849. 54. Turk, A., Sakalis, E., Rago, 0., and Karamitsos, H. (1992). Activated carbon systems for removal of light gases. Ann. N Y Acad. Sci., 661,221-8. 55. Boudou, J.P., Chehimi, M., Broniek, E., et al. (2003). Adsorption of H 2S or S02 on an activated carbon cloth modified by ammonia treatment. Carbon, 41, 1999-2007. 56. Yang, Q.H., Zheng, J.T., Li, Y., et al. (1999). Adsorption and conversion of hydrogen sulfide over PAN-based ACF. Carbon, 37, 2078-80. 57. Adib, F., Bagreev, A., and Bandosz, T.J. (2000). Adsorption/oxidation of hydrogen sulfide on nitrogen modified activated carbons. Langmuir, 16, 1980-6. 58. Kastner, J.R., Das, K.C., and Melear, N.D. (2002). Catalytic oxidation of gaseous reduced sulfur compounds using coal fly ash.]. Hazard. Mater., B95, 81-90.
560
Chapter
21
Removal of Inorganic Gases and VOCs on Activated Carbons
59. Kastner, J.R. Das, K.C., Buquoi, Q., and Melear, N.D. (2003). Low temperature catalytic oxidation of hydrogen sulfide and methanethiol using wood and coal fly ash. Environ. Sci. Techno!., 37, 2568-74. 60. Cal, .M.P., Strickler, B.W., and Lizzio, A.A. (2000). High temperature hydrogen sulfide adsorption on activated carbon. I. Effect of gas composition and metal addition. Carbon, 38, 1757-65. 61. Bansal, R.C., Donnet,J.B., and Stoeckli, F. (1988). Active Carbon. Marcel Dekker. 62. Salame, 1.1. and Bandosz, T.J. (1999). Study of water adsorption on activated carbons with different degrees of surface oxidation J. Colloid Interface Sci., 210, 367-74. 63. Salame, 1.1. and Bandosz, T.J. (1999). Revisiting the effect of surface chemistry on adsorption of water on activated carbons J. Phys. Chern., 103, 3877-84. 64. Bagreev, A., Bashkova, S., Locke, D.C., and Bandosz, T.J. (2001). Sewage sludge derived materials as efficient adsorbents for removal of hydrogen sulfide. Environ. Sci. Technol., 35, 1537-43. 65. Bandosz, T.J. (1996-1999). Study of granular activated carbon: Progress report. Fall 1996-FaIl1999. NYC DEP. 66. Kohl, A. and Riesenfeld, F. (1985). Gas Purification, 4th edn. Gulf Publishing Company. 67. Stirling, D. (2000). The Sulfur Problem: Cleaning up Industrial Feedstocks. The Royal Society of Chemistry. 68. Rodriguez-Mirasol, J., Cordero, T., and Rodriguez, J.J. (1997). Effect of oxygen on the adsorption of S02 on activated carbon. Abstract of 23rd Biennial Conference on Carbon, July 18-23, College Park, p. 376. 69. Moreno-Castilla, C., Carrasco-Marin, F., Utrera-Hidalgo, E., and Rivera-Utrilla, J. (1993). Activated carbons as adsorbents of sulfur dioxide in flowing air. Effect of their pore texture and surface basicity. Langmuir, 9, 1378-83. 70. Lisovskii, A., Shter, G.E., Semiat, R., and Aharoni, C. (1997). Adsorption of sulfur dioxide by active carbon treated by nitric acid: II Effect of preheating on the adsorption properties. Carbon, 35, 1645-8. 71. Davini, P. (2001). S02 adsorption by activated carbons with various burnoffs obtained from bituminous coal. Carbon, 39, 1387-93. 72. Mochida, I., Miyamoto, S., Kuroda, K., et al. (1999). Adsorption and adsorbed species of S02 during its oxidative removal over pitch-based activated carbon fibers. Energy Fuels, 13, 369-73. 73. Davini, P. (1990). Adsorption and desorption of S02 on active carbon: the effect of surface basic groups. Carbon, 28, 565-71. 74. Lisovskii, A., Semiat, R., and Aharoni, C. (1997). Adsorption of sulfur dioxide by active carbon treated by nitric acid: I Effect of the treatment on adsorption of S02 and Extractability of the acid formed. Carbon, 35, 1639-43. 75. Anurov, C.A. (1996). Physicochemical aspects of the adsorption of sulfur dioxide by carbon adsorbents. (Uspekhi Khimit) Russ. Chern. Rev., 65, 663-76. 76. Raymundo-Pinero, E., Cazola-Amoros, D., Salinas-Martinez de Lecea, C., and Linares-Solano, A. (2000). Factors controlling the S02 removal by porous carbons: relevance of the S02 oxidation steep. Carbon, 38, 335-44. 77. Molina-Sabio, M., Munecas, M.A., Rodriguez-Reinoso, F., and McEnaney, B. (1995). Adsorption of CO 2 and S02 on activated carbons with a wide range of micropore size distribution. Carbon, 33, 1777-82.
References
78. Daley, M.A., Mangun, C.L., DeBarr,J.A., et al. (1997). Adsorption of50 2 onto oxidized and heat-treated activated carbon fibers (ACF5). Carbon, 35, 411-17. 79. Lizzio, A.A. and DeBarr, J .A. (1997). Mechanism of 502 removal by carbon. Enetgy Fuels, 11,284-91. 80. Rubio, B. and Izquierdo, M.T. (1997). Influence oflow-rank coal char properties on their S02 removal capcity from flue gases: I non-activated chars. Carbon, 35, 100-11. 81. Rubio, B., Izquierdo, M.T., and Mastral, A.M. (1998). Influence of low-rank coal char properties on their S02 removal capacity from flue gases.2. Activated chars. Carbon, 36, 263-8. 82. Mochida, I., Korai, Y., Shirahama, M., et al. (2000). Removal of SOx and NO x over activated carbon fibers. Carbon, 38, 22-29. 83. Davini, P., Stoppato G. (1997). S02 adsorption on active carbons: the effect of certain metal compounds. Abstract of 23rd Biennial Conference on Carbon, July 18-23, College Park, p. 316. 84. Roman, M.C., Takarada, T., Suzuki, Y., Linares, A. (1997). S02 interaction with a Ca-exchanged-coal. Abstract of23rd Biennial Conference on Carbon, July 18-23 College Park, p. 324. 85. Bagreev, A., Bashkova, S., and Bandosz, T.J. (2002). Adsorption of S02 on activated carbons: the effect of nitrogen functionality and pore sizes. Langmuir, 18, 1257-64. 86. Wang, Z-M. and Kaneko, K. (1998). Effect of pore width on micropore filling mechanism of S02 in carbon micropores. J. Phys. Chern. B, 102, 2863-8. 87. Lu, G.Q. and Do, D.D. (1993). Retention of sulfur dioxide as sulfuric acid by activated coal reject chart. Sep. Technol., 3, 106-10. 88. Bashkova, S., Bagreev, A., Locke, D.C., and Bandosz, T.J. (2001). Adsorption of S02 on sewage sludge-derived materials, Environ. Sci. Technol., 35, 3263-9. 89. Ho, C-S., Shih, S-M. (1992). Ca(OH)2/fly ash sorbents for S02 removal. Ind. Eng. Chern. Res., 31, 1130-5. 90. Leon y Leon, C.A. and Radovic, L.R. (1992). Interfacial chemistry and electrochermistry of carbon surfaces. In Chemistry and Physics of Carbon, Vol. 24 (P.A. Thrower, ed.). Marcel Dekker, pp. 213-310. 91. Yang, F.H. and Yang, R.T. (2003). An initio molecular orbital study of the mechanism of S02 oxidation catalyzed by carbon. Carbon, 41, 2149-58. 92. Kawabuchi, Y., Sotowa, C., Kuroda, K., et al. (1996). Preparation of active carbon fiber with basic properties. Abstracts, International Conference on Carbon, Carbon, Newcastle. UK, p. 431. 93. Lee,J.K., Shim, H.J., LimJ.C., et al. (1997). Influence of tension during oxidative stabilization on S02 adsorption characteristics of polyacrylonitrile (PAN) based activated carbon fibers. Carbon, 35, 837-43. 94. Alves, B.R. and Clark, A.J. (1986). An examination of the products formed on reaction of hydrogen cyanide and cyanogens with copper, chromium (6+) and copper-chromium (6+) impregnated activated carbons. Carbon, 24, 287-94. 95. Alder, J.F., Fielden, P.R., and Smith, S.J. (1988). The adsorption of hydrogen cyanide by impregnated activated carbon cloth. Part I: studies on cobalt and nickel acetatates as impregnants for hydrogen cyanide removal. Carbon, 25, 701-11. 96. Alder, J.F., Fielden, P.R., and Smith, S.J. (1988). The adsorption of hydrogen cyanide by impregnated activated carbon cloth. Part II. Reactivity of impregnated metal carboxylates towards hydrogen cyanide. Carbon, 26, 713-21.
562
Chapter
21
Removal of Inorganic Gases and VOCs on Activated Carbons
97. Raj akovic, Lj.V., Hic, M.R., Jovanic, P.B., and Radosevic, P.B. (1995). Stoichiometric analysis of chemisorption of hydrogen-cyanide onto activated carbon cloth. Carbon, 33, 1433-41. 98. Hall, P.G., Gittins, P.M., Winn, J.M., and Robertson, J. (1985). Sorption of phosphine by activated carbon cloth and the effects of impregnation with silver and copper nitrates and the presence of water. Carbon, 23, 353-71. 99. Singh, B., Prasad, G.K., Suryanarayana, M.V.S., and Banerjee, S. (2001). The reaction of thiodiglycol on metal-impregnated carbon. Carbon, 39, 2131-42. 100. Freeman, F.G. and Reucroft, P.J. (1979). Adsorption of HCN and H 20 vapor mixtures by activated and impregnated carbons. Carbon, 17, 313-16. 101. Rubel, A.M and Stencel, J.M. (1996). Effect of Pressure on NO x adsorption by activated carbons. Energy Fuels, 10, 704-70. 102. Izquierdo, M. T. and Rubio, B. (1998). Influence of char physicochemical features on the flue gas nitric oxide reduction with chars. Environ. Sci. Technol., 32, 4017-22. 103. Xia, B., Phillips, J., and Chen, C-K. (1999). Impact of pretreatments on the selectivity of carbon for NO x adsorption/reduction. Energy Fuels, 13, 903-6. 104. Kong, Y. and Cha, C-Y. (1996). NO x abatement with carbon adsorbents and microwave energy. Energy Fuels, 9, 971-5. 105. Kong, Y. and Cha, C-Y. (1996). Microwave-induced regeneration of NOx-saturated char. Energy Fuel, 10, 1245-9. 106. Lee, Y-W., Park,J-W., Yun,J-H., et al. (2002). Studies on the surface chemistry based on competitive adsorption ofNO x -S0 2 onto a KOH impregnated activated carbon in excess 02' Environ. Sci. Techno!', 36, 4928-35. 107. Lee, Y-W., Choi, D-K., and Park, J-W. (2001). Surface chemical characterization using AES/SAM and ToF-SIMS on KOH-impregnated activated carbon by selective adsorption ofNO x ' Ind. Eng. Chern. Res., 40, 3337-45. 108. Ulan-Gomez, MJ., Salina-Martinez de Lecea, C., Linares-Solano, A., and Radovic, L.R. (1998). Potassium-containing coal chars as catalysts for NO x reduction in the presence of oxygen. Energy Fuels, 12, 1256-64. 109. Garcia-Garcia, A., Illan-Gomez, MJ., Linares-Solano, A., and Salinas-Martinez de Lecea, C. (2002). NO x reduction by potassium-containing coal briquettes. Effect of preparation procedure and potassium content. Energy Fuels, 16, 569-74. 110. Garcia-Garcia, A., Illan-Gomez, MJ., Linares-Solano, A., and Salinas-Martinez de Lecea, C. (1999). NO x reduction by potassium-containing coal briquettes. Effect of N0 2 concentration. Energy Fuels, 13, 499-505. 111. Bueno-Lopez, A., Caballero, J.A., and Garcia-Garcia, A. (2002). Analysis of the reaction conditions in the NO x reduction process by carbon with a view to achieve high NO x conversions. Residence time considerations. Energy Fuels, 16, 1425-8. 112. Mochida, I., Korai, Y., and Shirahama, M., et al. (2000). Removal of SOx and NO x over activated carbon fibers. Carbon, 38, 227-39. 113. Matzer, S. and Boehm, H.-P. (1998). Influence of nitrogen doping on the adsorption and reduction of nitric oxide by activated carbons. Carbon, 36, 1697-709. 114. Oark, B.-J., Park, S-J., and Ryu, S-K. Removal of NO over copper supported on activated carbon prepared by electroless plating. J. Colloid Interface Sci., 217, 142-5. 115. Davini, P. (2001). S02 and NO x adsorption properties of activated carbons obtained from a pitch containing iron derivatives. Carbon, 39, 2173-9.
References
116. Yamshita, H., Tomita, A., Yamada, H., et al. (1993). Influence of char surface chemistry on the reduction of nitric oxide with chars. Energy Fuels, 7, 85-91. 117. Ulan-Gomez, M.J., Linares-Solano, A., Radovic, L.R., and Salina-Martinez de Lecea, C. (1996). NO reduction by activated carbons. 7. Some mechanistic aspects of uncatalyzed and catalyzed reaction. Enetgy Fuels, 10, 158-68. 118. Illan-Gomez, M.J., Linares-Solano, A., Radovic, L.R., and Salina-Martinez de Lecea, C. (1995). NO eeduction by activated carbons. 2. Catalytic effect of potassium. Enetgy Fuels, 9, 97-103. 119. www.epa.gov/ttn/atw/188polls.html 120. Yu, F.D., Luo, L.A., and Grevillot, G. (2002). Adsorption isotherms of VOCs onto an activated carbon monolith: experimental measurement and correlation with different models.]. Chem. Eng. Data, 47, 467-73. 121. Dubinin, M.M. (1966). Porous structure and adsorption properties of active carbons, In Chemistry and Physics of Carbon, Vol. 2 (P.L. Walker, ed.). Marcel Dekker, pp. 51-120. 122. Kiselev, A.V. and Yashin, Y.I. (1969). Gas Adsorption Chromatography, Plenum. 123. Bansode, R.R., Losso, J.N., Masrshall, W.E., et al. (2003). Adsorption of volatile inorganic compounds by pecan shell- and almond shell-based granular activated carbons. Bioresource Technol., 90, 175-84. 124. Pre, P., Delage, F., Faur-Brasquet, C., and Le Cloirec, P. (2002). Quantitative structure-activity relationship for the prediction ofVOCs adsorption and desorption energies onto activated carbon. Fuel Process Technol., 77-8, 345-51. 125. Chiang, Y-C., Chiang, P.C, and Chang, E.E. (2001). Effects of surface characteristics of activated carbons on VOC adsorption.]. Environ. Eng., 127, 54-62. 126. Delange, F., Pre, P., and Le Cloirec, P. (2000). Mass transfer and warming during adsorption of high concentrations of VOCs on an activated carbon bed: experimental and theoretical analysis. Environ. Sci. Technol., 34, 4816-21. 127. Do, D.D. and Hu. X. (1993). An energy-distributed model for adsorption kinetics in large heterogeneous microporous particles. Chem. Eng. Sci., 48, 2119-27. 128. Chou, M-S. and Chiou, J-H. (1997). Modeling effects of moisture on adsorption capacity of activated carbon for VOCs.]. Environ. Eng., 1213, 437-43. 129. Shin, H-C., Park,]-W., Park, K., and Song, H-C. (2002). Removal characteristics of trace compounds of landfill gas by activated carbon adsorption. Environ. Pollut., 119,227-36. 130. Domingo-Garcia, M., Fernandez-Morales, I., Lopez-Garzon, F.J., et al. (1999). On the adsorption of formaldehyde at high temperatures and zero surface coverage. Langmuir, 15,3226-31. 131. EI-Sayed, Y. and Bandosz, T.J. (2001). A study of acetaldehyde adsorption on activated carbon. J. Colloid Interface Sci., 242, 44-51. 132. EI-Sayed, Y. and Bandosz, T.J. (2002). Acetaldehyde adsorption on nitrogencontaining activated carbons. Langmuir, 18, 3213-18. 133. Fuertes, A.B., Marban, G., and Nevskaia, D.M. (2003). Adsorption of volatile organic compounds by means of activated carbon fibre-based monoliths. Carbon, 41,87-96. 134. Perez-Mendoza, M., Domingo-Garcia, M., and Lopez-Garzon, F.J. (2000). Adsorption of methylamines on carbon materials at zero surface coverage. Langmuir, 16,7012-18. 135. Abe, M., Kawashima, K., Kozawa, K., et al. (2000). Amination ofactivated carbon and adsorption characteristics of its aminated surface. Langmuir, 16, 5059-63.
Chapter
21
Removal of Inorganic Gases and VOCs on Activated Carbons
136. EI-Sayed, Y. and Bandosz, T J. (2003). Effect of increased basicity of activated carbon surface on valerie acid adsorption from aqueous solution. Phys. Chern. Chern .Phys., 5, 4892-9. 137. EI-Sayed, Y. and Bandosz, TJ. (2004). Adsorption of valerie acid from aqueous solutions on activated carbons; role of surface basic sites. J. Colloid Interface Sci. 138. Gironi, F., Capparucci, C., and Marrelli, L. (2003). Adsorption ofMTBE vapors onto activated carbon. J. Chern. Eng. Data, 48, 783-8. 139. Li, L., Quinlivan, P.A., and Knappe, D.R.U. (2002). Effect of activated carbon
140. 141. 142. 143.
surface chemistry and pore structure on the adsorption of organic contaminants from aqueous solution. Carbon, 40, 2085-100. Salame, 1.1. and Bandosz, TJ. (2000). Adsorption of water and methanol on micro- and mesoporous wood-based activated carbons. Langmuir, 16, 5435-40. Salame, 1.1. and Bandosz, TJ. (2001). Study of diethyl ether adsorption on activated carbons using IGC at finite concentration. Langmuir, 17, 4967-72. Hsieh, C.-T. and Chen, J.-M. (2002). Adsorption energy distribution model for VOCs onto activated carbons, J. Colloid. Interface Sci., 255, 248-53. Popescu, M., Joly, J.P., Carre, J., and Danatoiu, C. (2003). Dynamical adsorption and temperature-programmed desorption of VOCs (toluene, butyl acetate and butanol) on activated carbons, Carbon, 41, 739-48.
GAS SEPARATION AND STORAGE BY ACTIVATED CARBONS Shivaji Sircar Department of Chemical Engineering, Lehigh University, Bethlehem, PA, USA
Contents Introduction Activated Carbons for Gas Separation and Purification 22.3 Mechanisms of Gas Separation by Activated Carbons 22.4 Examples of Gas Separation Processes 22.5 Adsorptive Process Design 22.6 Storage of Natural Gas on Activated Carbons 22.7 Conclusions References
22.1
22.2
22.1 INTRODUCTION
Separation and purification of gas mixtures by selective adsorption of one or more components of the mixture on a micro- and meso-porous adsorbent is a major unit of operation in the chemical, petrochemical, environmental, and pharmaceutical industries. The phenomenal growth in the development of this technology during the last 20 years is demonstrated by Fig. 22.1. It shows the number of the US Patents issued every year between 1980 and 2000 on "gas separation by adsorption" and "adsorption for air pollution control." More than Adsorption by Carbons ISBN: 978-0-08-044464-2
© 2008 Elsevier Ltd.
All rights reserved.
565
566
Chapter
Gas Separation and Storage by Activated Carbons
22
US Patents (1980-2000)
250
r---------'----------,
200
Co
Q)
150
VI
tu
G")
N N
~
"0
tu
:::r
()
o
00
U"1
22.4
Examples of Gas Separation Processes
Table
22.7
O 2 (1)
581
Separation performance of MSC membranes
+ N2
(2)
H 2 (1) + CH 4 (2) CO 2 (1) + CH4 (2) C 3 H 6 (1) + C 3 H s (2)
16.0 365.5 91 183
13,6 500 50 12-15
[23] [22] [22]
[22]
pressure from a gas contaInIng 50 % H 2 + 50 % H 2 S [25]. This remarkable separation is achieved by operating the first stage of the membrane with a low H 2 recovery (r-v30 %) and the second stage with a high H 2 recovery (r-v90 %) [25]. The net H 2 S rejection by the membrane is 98.3 %. The H 2 can be further enriched to make a 99.99+ % pure H 2 product by using a conventional TSA system using an activated carbon adsorbent. The overall H 2 recovery by the SSFTSA hybrid process is r-v77 %. The SSF membrane has been scaled up and pilot tested [31].
22.4.5 Sorption-Reaction Process for Removal of Trace VOC Stricter environmental regulations often require thermal or catalytic incineration oftrace hydrocarbons at a temperature of600-1600 K from contaminated air before venting [32]. This usually consumes a large quantity of fuel (energy) to heat the air mass. A cyclic sorption-reaction (SR) process concept developed by the Air Products and Chemical, Inc., USA, can reduce the energy need for this application [33]. The process consists of (i) adsorption of trace hydrocarbons from the air at ambient temperature and pressure in a bed packed with a mixture of an activated carbon and an oxidation catalyst to produce a clean air stream, (ii) in situ oxidation of the hydrocarbons by directly or indirectly heating the packed bed to about 425 K, and (iii) cooling the bed to near ambient temperature and venting the combustion products. The cycle is then repeated. Only the adsorber vessel and the adsorbent-catalyst mixture are heated to the reaction temperature which reduces the energy need. The adsorption of trace hydrocarbons and their subsequent batchwise thermal desorption substantially increase their concentrations in the gas phase which facilitate the reaction rates. Figure 22.8(a) shows the schematic drawing of a two-column embodiment of the SR process. Figure 22.8(b) shows a shell and tube reactor design for the process using indirect heating and cooling during the cycle. Table 22.8 compares the performance of the SR process for cleaning a 1 MM SCFD (million standard cubic feet per day) (0.0283 x 106 m 3 / day) air stream containing 260 ppm vinyl chloride monomer (VCM) to a level of 1 ppm with that of a conventional plug-flow reactor using a standard oxidation catalyst at a reaction temperature of 600 K [18]. The adsorbent in the SR process was RB
582
Chapter
22
Gas Separation and Storage by Activated Carbons
Time (h) To vent
o 20 1A_A_ H
40
lei
leMA_ 1A
1B
Adsorbent and catalyst Clean air
..........
.....---~
"--"'I~I"""'----.l----"Clean
Blower
Blower
Air -+- )
air Heating fluid
Heating fluid
(a)
(b) 0.3
,.-------------------"'!"'l • Experiment p= 1.0 atm -
t
Langmuir model
0.2
100
200
y(ppm) --...
(c)
Figure 22.8 Sorption-reaction (SR) process for removal of trace organic contaminants: (a) schematic drawing of a two column SR process, (b) shell and tube reactor configuration for the process, (c) isotherms for adsorption of trace vinyl cWoride monomer (VCM) on an activated carbon.
Table 22.8
Energy savings by sorption - reaction process
Power requirement (kW) Adsorbent - catalyst inventory (kg)
3.52 2590
120 364
22.4
Examples of Gas Separation Processes
carbon that was impregnated with 1.5 wt % palladium chloride as the oxidation catalyst. Figure 22.8(c) shows the isotherms for adsorption of trace VCM from air on the RB carbon at three temperatures. The table shows that energy savings of an order of magnitude can be achieved by the SR process.
22.4.6 Chemically Modified Activated Carbons for Gas
Separation Molecular engineering of an activated carbon surface provides a very interesting opportunity to beneficially alter the carbon's core adsorptive characteristics for gas separation. Polar groups can be introduced to the surface of a weakly polar carbon by judicious surface oxidation. Thus, a hydrophobic carbon can be converted to a hydrophilic adsorbent. Figure 22.9(a) shows the water vapor adsorption isotherms at 297 K (specific amount of water adsorbed 11 as a function of relative water vapor pressure x) on the original Ceca carbon (Type V
100
[1'1'
,.
1""'1
99
Water adsorption isotherms at 24 0 C 0.3·
98 97
~ >.
~ ~
0.2
.9
o()
96 95
~
94
« '-
93 92
0.1
91 90
o
0.2
0.4
0.6
0.8
(a)
r'\'
"
i\
r\
.
;
\
\
\
M !\
~. :~~~~~j~~g:~5 40
il1\ •
J
60
50
1.0
x-
"-
70
80
90
100
H20 rejection (%)
(b)
0.35....------------------, - Y - MSC-30 -x- NO.5 0.30 - - NO.7 -a->- NO.8 0.25
~
0.20
(5
E E
o
0.15 0.10 0.05
o.00
~-l...--\.-__J.___J.._-J--...1,.".......JI..___.1.___1.._.I.__.L_,.....L_...L._...1.._J....,_1--l..:...-...L--J
0.0
(c)
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Pressure (bar)
Figure 22.9 Examples ofmolecular engineering of carbon surface: (a) isotherms for adsorption of H 2 0 vapor on various chemically modified carbons, (b) gas drying characteristics of the modified selective surface membrane (SSF) membrane, (c) high-temperature chemisorption of CO 2 on MgO-doped activated carbon.
Chapter
22
Gas Separation and Storage by Activated Carbons
isotherm shape by Brunauer classification [19]) and on carbon samples which have been oxidized by heating in fuming HN0 3 and 45 % HN0 3 solution with or without the presence of copper acetate catalyst at 353 K [34]. Oxidation progressively changed the shape of the water isotherms from Type V to Type IV to Type I indicating that the surface has changed from hydrophobic to mildly hydrophilic. The figure also gives the water adsorption isotherms on NaX zeolite (Type I with very strong adsorption) and Alumina (Type IV with moderately strong adsorption). The Type I water isotherm on the oxidized carbon produced by HN0 3 oxidation in presence of the catalyst exhibits higher water capacity than the alumina in the low-pressure region but it holds water much less strongly than the zeolite. The Henry's law constants for water adsorption on the NaX zeolite, the modified Ceca carbon, and the alumina at 297 K are ~140, 5.3, and 2.1 kg/kg, respectively [34]. These properties are very desirable for a desiccant to be used in a PSA gas drying process (moderate water adsorption capacity and ease of desorption by partial pressure reduction). The modified activated carbon can serve as an improved desiccant for a PSA drier. The carbon surface oxidation procedure described above was also applied to introduce hydrophilicity to the pore walls of the SSF membrane [25]. The oxidized membrane could remove water vapor from air at a moderate feed gas pressure of ~305-405 kPa. Water was selectively adsorbed and permeated through the membrane. The adsorbed water essentially blocked the flow of air molecules through the void space of the modified membrane. Figure 22.9(b) shows the performance of the modified SSF membrane. Air recovery (fraction of air feed that is produced as the high-pressure product gas) is plotted against water rejection (fraction of feed water that is permeated to the low-pressure side) in the figure. The low-pressure side of the tubular SSF membrane was purged with N 2 . The parameter S/F represents the ratio of purge gas flow rate to that of feed gas. It can be seen that 70-80 % of feed water can be rejected with very little air loss (~99 % of feed air is recovered at the feed pressure). The original membrane was totally unselective toward water. The modified SSF membrane can be used as a continuous drying apparatus. Activated carbon surfaces have also been modified by doping them with various metal oxides, such as CaO or MgO or their combinations, in order to produce chemisorbents which are selective toward CO 2 at high temperatures (600-700K) [35]. Figure 22.9(c) shows isotherms for adsorption of CO 2 at 573 K on various doped samples of Maxsorb carbon (MSC-30) produced by the Kansai Coke and Chemical Company ofJapan [35]. Of particular interest is the isotherm No.6 which is a MgO-doped carbon. It exhibits large CO 2 working capacity. Reversible high-temperature CO 2 chemisorbents can be used in the newly developed "Sorption Enhanced Reaction Process" by the Air Products and Chemicals, Inc., USA. The process can produce COx-free H 2 (fuel cell grade) by carrying out the equilibrium-controlled steam methane reforming reaction at a much lower temperature than the conventional processes without sacrificing the conversion of CH 4 to H 2 [36].
22.5
Adsorptive Process Design
585
The above examples demonstrate that activated carbons can be chemically modified for different new gas separation applications and the possibilities are many.
22.5 ADSORPTIVE PROCESS DESIGN The state of the art procedure for design of cyclic PSA or TSA processes using activated carbon adsorbents is to simultaneously solve the partial differential equations describing the mass, the heat, and the momentum balance equations for each step ofthe process using the appropriate initial and boundary conditions. These numerical calculations are carried out over many cycles for the process until a cyclic steady-state performance solution is achieved. Many different numerical integration algorithms are available for this purpose. The core input variables for the solution are multicomponent gas adsorption equilibria, heats, and kinetics for the system of interest [37]. The separation of20 % C 2 H 4 (1) + 80 % CH 4 (2) gas mixture was mathematically simulated using the four-step Skarstrom PSA cycle consisting of adsorption, depressurization, product purge, and product pressurization steps [37]. The feed gas pressure and temperature were 2.08 MPa and 298 K, respectively. The Nuxit charcoal was employed as the adsorbent. An essentially pure stream of CH 4 product (99.87 %) could be produced with a recovery of 28.4 %. The CH 4 productivity was 0.088 kg moles of product/kg of carbon in system/day [37]. Figure 22.10(a) shows the profiles of gas phase mole fractions of C 2 H 4 inside the adiabatic adsorber as functions of column positions at the end of each step of the PSA cycle. Figure 22.10(b) shows the corresponding C 2 H 4 loading profiles on the carbon normalized to the C 2 H 4 capacity at the feed gas conditions. The shaded area in the figure represents the net cyclic working capacity for C 2 H 4 as a function of the column position. These profiles demonstrate the complexity of the process created by nonlinearity of the adsorption equilibria, adsorbent heterogeneity, and column nonisothermality. Most of the working capacity of the carbon for C 2 H 4 is located in the first half of the column (feed end). The second half of the column is used to achieve the high purity of the product gas. These results may not be intuitive. Table 22.9 compares the performance of the above-described PSA process for two extreme cases: (a) isothermal operation and (b) assumption of constant heats of adsorption for the components or ignorance of adsorbent heterogeneity. The severe influence of the energetic heterogeneity of the carbon and the corresponding thermal effects on the separation process is self-evident.
Chapter 22 Gas Separation and Storage by Activated Carbons
586
0.40
c
r-....._--r--__r_--r-....--.--_._--......-..ro--'t"'-"
r-
-_
(i) Repressurize (ii) Adsorb - - _. (iii) Depressurize (iv) Purge
0.30
0
U jg Q)
(5
E 0.20 Q)
c
Q)
~
.c [j
0.10
0.2
0.4
0.6
0.8
1.0
Normalized column axial position, Z*
(a)
iCc: ..... c:
1.00
0> c
:.cco .Q
(i) Repressurize (ii) Adsorb (iii) Depressurize (iv) Purge
0.75
Q)
c
Q)
~
.c Ci5
"0
0.50
Q)
.~
co
E
o
z
0.25
0.00 '--""-""'-....a--"""'.......r..--'--l""-I--""---'---t-~_..a..__L__t'--J..._""__~ . . .iiiilllI 1.0 0.2 0.4 0.6 0.8 0.0
Normalized column axial position, Z*
(b)
Figure 22.10 Adsorption column profiles for C 2 H 4-CH 4 separation by a pressure swing adsorption (PSA) process using activated carbon: (a) gas phase C 2 H 4 mole fractions, (b) C 2 H 4 loadings on the adsorbent.
22.6
Storage of Natural Gas on Activated Carbons
Table 22.9
Performance of the PSA process for (2 H4 + (H 4 separation
CH 4 purity (mol %) CH4 recovery (%) CH 4 productivity (kg moles/kg/day)
99.87 28.4 0.088
99.87 60.16 0.376
99.88 18.92 0.057
22.6 STORAGE OF NATURAL GAS ON ACTIVATED CARBONS The abundance of natural gas, its relatively lower price, and its potential to be a cleaner fuel has promoted much interest in its use as motor fuel. Numerous vehicles around the world have been adapted to use "compressed natural gas (CNG)" as fuel. The gas is typically stored at a pressure of200 atm (r-v20.2 MPa) in heavy steel cylinders. The net deliverable capacity for the CNG tank between 20.2 and 0.137 MPa is 215 standard literslliter of storage volume (m3 /m 3 ). The energy density of CNG is, however, only 29 % of that of gasoline. Replacing CNG by storing methane in a vessel packed with an activated carbon has been a subject of much work during the last 20 years. This concept of using "Adsorbed Natural Gas (ANG)" can potentially reduce the gas storage pressure without sacrificing deliverable capacity. The target pressure is r-v35 atm (r-v3.5 MPa) that can be obtained by a single-stage compressor. The key question is whether the deliverable capacity of ANG can match that ofCNG. Research in this area has been directed toward (a) finding or preparing activated carbons with high micropore volume (or BET surface area) in order to increase the CH 4 adsorption capacity, and (b) to increase the bulk density of the carbon (reduce void volume of the storage vessel) to minimize the amount of unadsorbed CH 4 in the tank. Chemical activation of carbon precursors by heating in alkali solutions and phosphoric acid to create micropores as well as surface activation of microporous carbons by heating in CO 2 and steam are common techniques used for this purpose [38, 39]. Table 22.10 summarizes the published values of isothermal deliverable CH 4 capacities for several activated carbon samples[18, 40-42]. The Amoco PX 21 carbon has the highest BET surface area of all carbons listed in Table 22.3. Yet, the isothermal deliverable CH 4 capacity for this carbon is about 100 m 3 /m 3 [18]. This is due to its low bulk density (high external void). If this carbon can be produced in a monolith form so that the external void in the ANG tank is negligible, then the isothermal deliverable capacity (r-v21 0 m 3 /m 3 ) approaches that of the CNG [18]. Recently, a carbon monolith was fabricated by Lozano-Castello and coworkers, but it produced an isothermal deliverable capacity of r-v126 m 3 /m 3 only [40]. The carbon temperature increases during filling the cylinder with methane and it decreases during methane discharge due to the release and consumption of
588
Chapter
22
Gas Separation and Storage by Activated Carbons
Table 22.10 Isothermal deliverable CH 4 capacities of activated carbons (CNG == 21 5 m3 1m 3 ). . .
I
. .
...
.
.
UU>U
I
Amoco PX 21 at 303 K Amoco PX 21 at 303 K Amoco PX 21 at 298 K Amoco PX 21 at 298 K Activated carbon fiber (C0 2 activated) Activated carbon
Granular (bulk density == 0.3 kg/I) Monolith (no external void) (bulk density == 1.08 kg/I) Granular Bonded with polymer
Monolith
82.0 (calculated) 210.0 (calculated) 101.0 144.0 143.0
[42]
126.0
[40]
[18] [18]
[41 ] [41]
the heats of ad(de)sorption, respectively. Instantaneous heat removal or supply from or to an ANG system may not be practically possible. Consequently, the real deliverable capacity of the activated carbon system may be much less than the isothermal numbers given by Table 22.10 if a rapid fill-up or supply of CH 4 is required. The adiabatic delivery capacity (ADC) of the carbon may be a more appropriate variable in absence of a heat supply/removal mechanism. Figure 22.11 (a) shows CH 4 adsorption isotherms on PX21 activated carbon at different temperatures [18]. The isosteric heat of adsorption of CH 4 (q) is 16.7kJ/mole. Figure 22.11(b) shows the CH 4 adsorption isobars at 35.0atm (3.54 MPa) and 1.35 atm (0.137 MPa) on the same carbon as well as the adiabatic fill-discharge operating lines, which have slopes equal to the ratio of the heat capacity of the adsorbent (Cp = 0.25cal/g/C (1.046k]/kg/K)) to the isosteric heat of adsorption of CH 4 . Accordingly, a temperature rise and a drop of 66 and 77 K from a base temperature of 303 K will occur during adiabatic filling and discharge of CH 4 , respectively, when operated between these two pressure levels. The corresponding ADC for the granular carbon will be ~36.5 m 3 /m 3 only [18]. Attempts have been made to incorporate phase change materials with the activated carbons inside the ANG tanks in order to remove (or supply) the heat from (or to) the carbon so that the carbon remains isothermal during the filldischarge process [41]. This, however, lowers the carbon inventory in the tank and increases system cost. Instantaneous heat transfer between the carbon and the phase exchange material may also not be practically feasible. Several theoretical studies of natural gas storage on activated carbons have been undertaken. A molecular simulation of CH 4 adsorption predicts that the maximum storage capacity by a palletized and a monolith activated carbon will be 146 and 209 m 3 1m3 , respectively [43]. Nonisothermal fill-discharge and dynamics ofCH 4 ad(de)sorption in ANG systems have also been evaluated [44].
22.7
Conclusions
100.0 - - - - - - - - - - - - - - - - - - - - - - . . . ,
Methane adsorption isotherms on PX 21 carbon
t
10.0
C)
en Q)
"'6 E
5
1.0
c::-
qO =4.0 kcal/mole
0.1
---------------"'--------1 1.0 10.0 100.0
0.1
P(atm)
(a)
---.......
r----------------------.. . . . . .
15
10
~ Q)
"'6
E
5 c:
5
I - ....1--1.. . . Slope =Cp t
P=1.35AI
o
-50
o
30
100
q 150
(b)
Figure 22.11 Adsorption of CH 4 on PX 21 activated carbon for storage: (a) adsorption isotherms at various temperatures, (b) adsorption isobars and operating lines for CH4 fill and discharge.
22.7 CONCLUSIONS Activated carbons produced from different precursors provide a large spectrum of pore structures and surface chemistry for gas separation and purification. The wide range of core adsorptive properties like adsorption equilibria, heats, and kinetics exhibited by these activated carbons encourages the design and
Chapter
59°
22
Gas Separation and Storage by Activated Carbons
development of novel pressure swing and temperature swing adsorption processes for gas separations. Many successful commercial applications in the areas of trace impurity removal from a contaminated gas, solvent vapor recovery, production of nitrogen from air, production of hydrogen and carbon dioxide from reformer off-gases, etc., already employ activated carbon adsorbents almost exclusively. Potential novel applications include nanoporous carbon membranes, combined sorption and reaction in a single unit operation, and natural gas storage. Molecular engineering of carbon pore structures and surface chemistry can open many new doors for application of this fascinating material in the gas separation industry.
REFERENCES 1. Sircar, S. (2000). Publications on adsorption science and technology. Adsorption, 6,359-65. 2. Sircar, S. (2001). Potential applications of gas separation by adsorption for the future. Adsorp. Sci. Tech., 19, 347-65. 3. Keller, G.E., Anderson, R.A., and Yon, C.M. (1987). Adsorption. In Handbook of Separation Process Technology (R.W. Rousseau, ed.). John Wiley, Chapter 12, pp. 644-96. 4. Lovett, W.D. and CunnifL F.T. (1974). Air pollution control by activated carbon. Chern. Eng. Prog., 70(5), 43-7. 5. Logsdon, P.B. and Basu, R.S. (1993). Recovery and recycle ofHCFCs by activated carbon adsorption. JIES, 36(2), 33-6. 6. Takeuchi, Y. (1991). Recent Advances in Solvent Recovery by Fixed-Bed Adsorption. Proceedings ofSeiken Symposium (M. Suzuki, ed.). Institute of Industrial Science, pp.87-94. 7. Schroter, H.J. andJiintgen, H. (1989). Gas separation by pressure swing adsorption using carbon molecular sieves. In Adsorption: Science and Technology, NA TO ASI Series, Vol. 158 (A.E. Rodrigues, M.D. Levan, and D. Tondeur, eds). Kluwer Academic, pp. 269-83. 8. Jiintgen, H., Knoblauch, K., Reichenburger, J., et al. (1981). Recovery of nitrogen rich gases from gases containing nitrogen and oxygen such as air. US Patent 4,263,339. 9. Knoblauch, K., Heimbach, H., and Harder, B. (1985). Preparation of nitrogen. US Patent 4,548,799. 10. Lacava, A.I. and LemcofL N.G. (1996). Single bed pressure swing adsorption process to generate high purity nitrogen. Gas. Sep. Purij., 10(2), 113-15. 11. Shirley, A.I and Lacava, A.I. (1992). Pressurization pressure swing adsorption (psa) systems for the production of high purity product gas. US Patent 5,082,474. 12. Lemco~ N.G. and Gmelin, R.C. (1993). Pressure swing adsorption method for separating gas mixtures. US Patent 5,176,722. 13. Golden, T.C. Battavio, P.J. Chen, Y.C., et al. (1993). Carbon-based oxygen selective desiccant for use in nitrogen PSA. Gas. Sep. Purij., 7, 274-8.
References
59 1
14. Litvinov, V.N. (1993). Removal of carbon dioxide from flue gases. Russian Patent 1,790,981. 15. Kapoor, A. and Yang, R.T. (1989). Kinetic separation of methanecarbon dioxide mixture by adsorption on molecular sieve carbon. Chern. Eng. Sci., 44, 1723-33. 16. Fuderer, A. and Rudelstorfer, E. (1976). Selective adsorption process. US Patent 3,896,849. 17. Sircar, S. (1979). Separation of multi-component gas mixtures. US Patent 4,171,206. 18. Sircar, S. Golden, T.C., and Rao, M.B. (1996). Activated carbon for gas separation and storage. Carbon, 34(1), 1-12. 19. Young, D.M. and Crowell, A.D. (1962). Physical Adsorption of Gases. Butterworths. 20. Skarstrom, C.W. (1960). Method and apparatus for fractionating gaseous mixtures by adsorption. US Patent 2,944,627. 21. Hirose, T. and Kuma, T. (1990). Honeycomb rotor continuous adsorber for solvent vapor recovery and dehumidification. 2nd Korea-Japan Symposium on Separation Technology. 22. Rudisill, E.N. Hacskaylo, J.J., and Levan, M.D. (1992). Co-adsorption of hydrocarbons and water on BPL activated carbon. I & E C Res., 31(4), 1122-30. 23. Rynders, R.M. Rao, M.B., and Sircar, S. (1997). Isotope exchange technique for measurement of pure and multi-component adsorption equilibria and kinetics. AIChEJ., 43, 2456-70. 24. Sircar, S. and Golden, T. C. (1995). Isothermal and isobaric desorption of carbon dioxide by purge. I & EC Res., 34, 2881-8. 25. Sircar, S. and Rao, M.B. (2000). Nano-porous carbon membranes for gas separation. In Recent Advances on Gas Separation by Micro-Porous Membranes (N. Kanellopoulos, ed.). Elsevier, pp. 473-6. 26. Carbon Membranes Ltd. Israel. Trade Literature. 27. Jones, C.W. and Koros, W.J. (1994). Carbon molecular sieve gas separation membrane: I - preparation and characterization based on polyamide precursors. Carbon, 32, 1419-25. 28. Tanihara, N.H., Shimazaki, Y., Hirayama, S., et al. (1999). Gas permeation properties of asymmetric carbon hollow fiber membranes prepared from asymmetric polyamide hollow fiber. J. Membrane Sci., 160(2), 179-86. 29. Rao, M.B. Sircar, S., and Golden, T.C. (1992). Gas separation by adsorbent membranes. US Patent 5,104,425. 30. Soffer, A. and Koresh, J.E. (1987). Separation device. US Patent 4,685,940. 31. Naheiri, T. Ludwig, K.A., Anand, M., et al. (1997). Scale-up of selective surface flow membrane for gas separation. Sep. Sci. Technol., 32, 1589-602. 32. Spivey, J.J. (1987). Complete catalytic oxidation of volatile organics. I & E C Res., 26, 2165-80. 33. Dalton, A.1. and Sircar, S. (1977). Method for removing low concentration of oxidizable organic contaminants from an oxygen containing inert gas. US Patent 4,025,605. 34. Golden, T.C. and Sircar, S. (1990). Activated carbon adsorbent for drying gases by pressure swing adsorption. Carbon, 28, 683-90. 35. Yong, Z., Mata, V.G., and Rodrigues, A.E. (2001). Adsorption of carbon dioxide on chemically modified high surface area carbon-based adsorbents at high temperature. Adsorption, 7, 41-50.
59 2
Chapter
22
Gas Separation and Storage by Activated Carbons
36. Waldron, W.E. Hufton, J.R., and Sircar, S. (2001). Production of hydrogen by cyclic sorption enhanced reaction process. AIChE J., 47, 1477-9. 37. Hartzog, D.G. and Sircar, S. (1995). Sensitivity of PSA process performance to input variables. Adsorption, 1, 133-51. 38. Ling, L. Song. Y. Liu, L., and Li. K. (2001). Method for preparing activated carbon for storing methane. Chinese Patent 1,303,732. 39. Baker, F.S. (1997). Low pressure methane storage with highly micro-porous carbons. US Patent 5,626,637. 40. Lozano-Castello, D., Cazoria-Amoros, D., Linares-Solano, A., and Quinn, D.F. (2002). Activated carbon monoliths for methane storage: Influence of binder. Carbon, 40, 2817-25. 41. Blazek, C.F., Jasionowski, W.J., Tiller, A.J., and Gauthier, S.W. (1990). Paper Presented at 25th Intersociety Energy Conversion Engineering Conference. Reno, Nevada. 42. Sejnoha, M., Chahine, R., Yaici, W., and Bose, T.K. (1994). Adsorption storage of natural gas. Paper presented at AIChE Annual Meeting, San Francisco, California. 43. Matranga, K.R., Myers, A.L., and Glandt, E.D. (1992). Storage of natural gas on activated carbon. Chern. Eng. Sci., 47, 1569-79. 44. Barbosa-Mota, J.P., Rodrigues, A.E., Saatdjian, E., and Tondeur, D. (1997). Dynamics of natural gas adsorption storage system employing activated carbon. Carbon, 35(9), 1259-70.
ELECTROCHEMICAL ENERGY STORAGE Fran~ois Seguin 1 and Elzbieta Frackowiak2 Centre de Recherche sur /a Matiere Divisee, CNRS-Universite, 1 B rue de /a Ferol/erie, 45071 Orleans Cedex 02, France 2 Institute of Chemistry and Technical Electrochemistry, Poznan University of Technology, ul. Piotrowo 3, 60-965 Poznan, Poland 1
Contents 23.1 Introduction 23.2 Lithium Insertion in Carbon Materials 23.3 Nanostructured Carbons as Components of Supercapacitor Electrodes 23.4 General Conclusion and Perspectives References
593 595 607 621 623
23.1 INTRODUCTION
The great demand for portable electronic devices supplied by light electric power sources as well as an increasing interest for electrically powered vehicles results in a continuous development of high-performance energy systems. Lithium-ion (Li-ion) and nickel-metal hydride batteries, fuel cells, and supercapacitors [1] belong to such promising energy sources. An intense research effort is focused on the development and/or improvement of electrode materials for these electrochemical systems. It is now well-demonstrated that carbons constitute a versatile class ofmaterials for the development of high-performance power sources [2, 3]. The important Adsorption by Carbons ISBN: 978-0-08-044464-2
© 2008 Elsevier Ltd.
All rights reserved.
593
594
Chapter 23 Electrochemical Energy Storage
factors that dictate the selection of carbon for this target are its accessibility, easy processability, and low cost. Moreover, the great versatility of the forms attainable allows to use carbons in composite electrodes or as a main electrode component when self-standing films, fabrics, felts, or aerogel blocks are used. Carbon as electrode is a well-polarizable material, chemically stable in different solutions (acidic, basic, aprotic) and in a wide range of temperatures. Because of its amphoteric character, carbon can be electron donor and acceptor as well, and its substitution by neighbor elements from the periodic chart (e.g., boron, nitrogen) allows the electronic properties to be adjusted, which broadens the panel of its electrochemical applications. Moreover, intercalation-insertion processes into carbon can proceed easily because of the weak van der Waals forces between graphitic layers and/or domains [4]. Hence, different carbons (e.g., graphites, cokes, mesophase, activated carbons, aerogels, xerogels, carbon blacks, nanotubes) have been extensively considered as active material for energy storage [5-8]. The nanotextural/structural and chemical properties of carbons determine their efficiency in electrochemical application as electrodes. A strict control of the carbonization process (time, temperature, gas flow), the kind of natural or synthetic precursor and/or chemical vapor deposition conditions allow to prepare carbons with almost defined structure/nanotexture. Various advanced forms of carbon can be designed and prepared by a careful selection of templates, in such a way that one-, two-, or three-dimensional carbons can be easily obtained [9-12]. The modification of carbon by an activation process gives a further possibility of affecting the properties, especially by high developing of the specific surface area [13]. The main demanded characteristics of carbon for all the electrochemical applications are good electrical conductivity and wettability, the latter being strongly affected by the surface functionality. The electrical conductivity significantly depends on the thermal treatment conditions, nanotexture, hybridization, and content of heteroatoms. Certainly, each electrochemical application involves completely different nanotextural properties of carbon characterized by the specific surface area, the presence of micro- and mesopores, their ratio, pores shape, and so on. Often literature claims about a correlation between some electrochemical parameters of a cell and the Brunauer, Emmett, Teller (BET) specific surface area. However, it is noteworthy that the electrochemically active surface area (ASA) of carbon, which takes part in the electrochemical processes, differs significantly from the physical surface area determined by nitrogen and/or carbon dioxide sorption. Therefore, the estimation of the ASA [14] connected with the special reactivity of carbon with di-oxygen can supply an additional and extremely useful information on the electrochemical properties of carbons in energy storage systems. In this chapter, the electrochemical applications of various carbon materials in energy conversion, mainly lithium storage and supercapacitors, will be critically discussed taking into account their structure/nanotexture and surface functionality, with some attention for future perspectives.
23.2
Lithium Insertion in Carbon Materials
595
23.2 LITHIUM INSERTION IN (ARBON MATERIALS The new generation ofLi-ion batteries combines a high power and energy density, hence, they appear as the most promising system for portable devices as well as for electrical vehicles. However, there is still a need to improve the electrode materials for getting the highest capacity, while keeping good electrochemical characteristics, especially a good calendar life. A relatively high reversible capacity (372mAh/g, i.e., one lithium for six carbon atoms in standard conditions) at a potential close to metallic lithium and a moderate irreversible capacity can be obtained with graphite-based anodes. A higher degree of reversible lithium insertion than in graphite, but also an important irreversible capacity, is observed with various kinds of nanostructured carbons. Therefore, an intensive research effort is focused on the optimization of the anodic carbon materials, with the objectives to enhance the reversible capacity and to reduce as much as possible the irreversible capacity and hysteresis, which are often important drawbacks of these materials. The next section will discuss the correlations between the electrochemical performance of nanostructured carbons and their nanotexture/structure and surface functionality. Taking into account the key parameters that control the electrochemical properties, some optimizations proposed in literature will be presented. 23.2.1
Principle of a Li-ion Battery
The electrodes of Li-ion batteries are based on intercalation materials between which lithium ions are transferred through an aprotic electrolytic medium during charge and discharge. The general principle of a Li-ion battery is presented in Fig. 23.1. The cathodic (positive electrode) materials are lamellar oxides such as LiCo0 2 and LiNi0 2 represented by the general formula Li yM0 2 (y ~ 1), whereas carbon (more generally graphite) is used for the negative electrode (anode). The following equations represent the reversible redox reactions in the cell:
6C + xLi+ + xe- {:} Li x C 6 Li y M0 2
{:}
Li y_x M0 2 + xLi+ + xe-
(23.1) (23.2)
Using graphite, such a battery operates at almost constant voltage of about 3.5 V during discharge, which makes this system very attractive for its highenergy density [15]. In the case of graphite, lithium penetrates between the graphene layers through an intercalation process with charge transfer to carbon. The successive formation of stages 3, 2, and 1 derivatives [16] during the reduction of graphite is demonstrated by well-defined plateaus at constant potential on the galvanostatic curves (Fig. 23.2), reaching the composition LiC 6 at saturation. It is remarkable for graphite that insertion and extraction proceed very close to the potential of metallic lithium, allowing the Li-ion battery to discharge at
Chapter 23 Electrochemical Energy Storage
~e-
~e-
+
•
Li+
~.
Li+
Electrolyte
Figure 23.1 Principle of a lithium-ion battery. This scheme shows the case of charge, i.e., graphite is reduced, while the oxide LiyMO z is oxidized. The part of lithium which is irreversibly lost during the first cycle is represented schematically as a layer surrounding graphite.
Carbon layer
--.--.. -••
Lithium 3
0.3
2.5
0.25
---
--.
••
0.2 2
Stage III
0.15 ::J
en >
1.5
--
----
••
•• ••
•• ••
••
••
Stage II
Sta el
0.1
~
0.05 0 0.2
0.5
0.4
0.6
1.2
0.8
x in Li xC6
o o
0.2
0.4
0.6
0.8
1.2
x in Li xC6
Figure 23.2 Galvanostatic intercalation-deintercalation of lithium in graphite using a twoelectrode lithium-graphite cell. The inset is a magnification of the curve at low potential, which shows the existence ofdifferent stage domains. The potential plateaus during reduction represent the successive transitions from stages III to II and to I.
23.2
Lithium Insertion in Carbon Materials
597
high and almost constant value of voltage. Nevertheless, it must be pointed out that a part of lithium, being involved in the so-called solid electrolyte interphase (SEI) [17] as LiF, Li2 0, LiOH, Li 2 C0 3 , ROC0 2 Li [18-20], formed during the first reduction cycle, is not recovered during deintercalation (oxidation) giving rise to a noticeable irreversible capacity. The SEI represented schematically in Fig. 23.1 as a layer surrounding graphite, is electron insulating, but behaves as an ionic conductor, allowing only nonsolvated lithium ions to migrate from the electrode surface to the bulk of graphite. The SEI formation is extremely profitable because it coats carbon preventing from further electrolyte decomposition on its surface [17]. Although graphite is the most used anodic material in commercial batteries, it has some drawbacks, especially because of the risks of exfoliation during cycling, which considerably affects the long-term life ofthe system. It is generally accepted that exfoliation is due to solvated lithium intercalation during the reduction step [17, 21]; therefore, it is important to control the initial formation of the SEI that is further supposed to allow only nonsolvated lithium to be intercalated.
23.2.2
Properties of Nanostructured Carbon Anodes
Lately, many efforts have been devoted to develop electrodes based on the use of hard carbons, because of the high reversible capacity which can be reached with some of these materials, without the inconvenience of exfoliation during the insertion-deinsertion processes. Beside these advantages, the drawbacks are an important irreversible capacity Cirr and a varying voltage during lithium deinsertion, giving rise to the so-called hysteresis. An example of charge-discharge characteristics of such a material in the form of carbon cloth from viscose is shown in Fig. 23.3 [22]. An optimal nanostructured carbon for Li-ion batteries should have a higher reversible capacity than graphite, while keeping a small irreversible capacity and its main part of discharge (oxidation) close to 0 V vs Li. Only a good knowledge of the physicochemical parameters that control the electrochemical properties of hard carbons can provide a chance to improve the materials and to reach ideal properties.
23.2.2.1
Origins of the irreversible capacity in nanostructured carbons
The possible origins ofirreversible capacity Cirr have been carefully studied by a number of authors. Some papers claim that lithium could be irreversibly trapped by the surface functional groups of carbon [23], e.g., by electrostatic forces such as -COO-Li+, or that it could react with di-oxygen or water molecules adsorbed on the carbon surface [24]. A linear dependence has been found between the irreversible capacity ofa series of carbons and their micropore volume [25], and 7Li NMR experiments lead to the conclusion that metallic
598
Chapter 23 Electrochemical Energy Storage
2.4
~--.,r------------------------.
2.2 2
0.71
1.34
1.8 :.:J
1.6
~
1.4 1.2
~
co ·E Q)
(5 (L
1
0.8 0.6 0.4 0.2
o -0.2 +&--&-.....I......I..~L...J..+.I....L.-I-...I....i-I-..L-L-L-+-L-I....&.......L-1f--L.L-.a........L.+.1..L-L...J....+-L....L...J-1-+-L-JL....L-L-f-J--l-.L.....L+...L..L-J...J...+-L...L..L.J-+-J-J~ -0.2 0 0.2 0.4 0.6 0.8 1.2 1.4 1.6 1.8 2 2.2 2.4
xin Li xC6
Figure 23.3 Galvanostatic insertion-deinsertion of lithium into hard carbon from viscose. Current load of20mA/g. (Adapted from Re£ [22]).
lithium could be irreversibly trapped in the micropores of the materials [26]. However, the main contribution to the irreversible capacity seems to be the formation of the SEI during the first reduction cycle at ca. 0.8 V vs Li. Some papers claim that the extent of the decomposition reaction is directly related to the BET specific surface area of carbon measured by nitrogen adsorption at 77 K [27-29]. In fact, this relationship between Cirr and the BET surface area is not always verified, as shown, e.g., in Fig. 23.4 for graphite ball-milled in different atmospheres and for the same samples coated by pyrolytic carbon after milling, confirming that other parameters control the value of C irr • This is not surprising, because the BET specific surface area is essentially a geometric parameter based on nitrogen physisorption on the basal carbon planes, whereas the SEI formation involves higher energies. Therefore, it has been suggested [22] to correlate the irreversible capacity to so-called ASA [14], which depends on the number of active sites on the carbon surface. The ASA of carbon materials corresponds to the cumulated surface area of the different types of defects present on the carbon surface (stacking faults, single and multiple vacancies, dislocations) [14, 30]; these sites are responsible for the interactions with the adsorbed species. A perfect linear relationship between the irreversible capacity and the value of ASA has been documented for different series of carbon samples [22]. While C irr can be possibly not correlated with the BET area, Fig. 23.4 shows that it is linearly dependent of the ASA [31]. Moreover, all the samples coated with a thin carbon layer by pyrolytic decomposition of propylene demonstrate the lowest values of irreversible capacity and ASA (Fig. 23.4) [22, 31]. Figure 23.5 illustrates the positive effect of such a coating on the charge-discharge characteristics of carbon fibers from viscose,
23.2
lithium Insertion in Carbon Materials
~
S
500
599
(9).
450
(I)
m 400 (ij 350
(a).
~ 300
(c).
«S 't: 250 ::J
en 200
(e).
(.)
~
(.)
150
~ 100
en ~
w
m
50
(h) (d) .(f) ,(b)
0 0
0.5
1.5
1
2.5
2
xin Li xC6 '@
70
N'
60
«S
50
S Q)
(ij Q) (.)
«S 't:
(c) (II)
•
(e) (9)
40
(a)
30
•
::J
en Q)
>
U «
20 10 0.5
1
1.5
2
xin Li xC6
Figure 23-4 Relation between the BET specific surface area (curve I) or the active surface area (ASA) (curve II) and the irreversible capacity x of graphite samples ball-milled in different conditions, or ball-milled and subsequently coated by pyrolytic carbon. (a) 10 h in vacuum; (b) 10 h in vacuum + pyrolytic carbon deposition; (c) 10 h under H 2 ; (d) 10 h under H 2+ pyrolytic carbon deposition; (e) 10 h under 02; (f) 10 h under 02+ pyrolytic carbon deposition; (g) 20 h in vacuum; (h) 20 h in vacuum + pyrolytic carbon deposition. (Adapted from Re£ [31].)
whose irreversible capacity diminishes to only x = 0.25 (x = 1 for a capacity C = 372 rnA h/g) , whereas before coating the same material showed a value x = 0.7 (Fig. 23.3) [22]. A similar improvement was mentioned after coating of graphite and subsequent carbonization, either with a blend ofpitch and resin [32] or with coal-tar pitch (CTP) [33]. Figure 23.6(a) shows the transmission electron microscopy (TEM) image of the cross section of a fiber coated with pyrolytic carbon. One can easily differentiate the fiber core with a highly disordered nanotexture from the more dense coating with a preferentially ordered nanotexture [22]. A schematic representation of this composite material is represented in Fig. 23.6(b). Because of its oriented nanotexture with only few edge planes accessible for adsorbed species, the thin carbon coating forms a barrier preventing an easy diffusion of the large solvated lithium ions to the active sites of the fibers. For the same reasons, in the ASA determination experiment, di-oxygen
600
Chapter 23 Electrochemical Energy Storage
2.4 . . . . - - . , - - - - - - - - - - - - - - - - - - - - - - - - - - , 2.2 2
0.25
1.51
1.8
:.:J 1.6 ~
~
1.4 1.2
(ij
1
Q)
0.8
E "0
a.
0.6 0.4 0.2
o -0.2 -t-L-'.....L.....J-~.L.......I..4_...L-L..L..L..f_L....&.......I.._+__L_'L-L-L-+_&__I_..L.....L..+~...L..+J.....L....J....L_+_'__.L-L...L..+_&__I_....L.....I..4...L....L..1.....L.+_JL-L-L-J....+_I_.L...L-L...I -0.2 o 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2.2 2.4 2
xin Li x C 6
Figure 23.5 Galvanostatic insertion-deinsertion of lithium in a composite constituted from hard carbon from viscose coated by a pyrocarbon film. Current load of 20 mAl g. (Adapted from Ref. [22].)
cannot easily diffuse to reach the active sites through the pyrolytic carbon layer, therefore the ASA is apparently low in the coated samples. 23.2.2.2
Properties of anodes from carbon nanotubes
Besides the typically microporous hard carbons, essentially mesoporous carbons such as nanotubes and nanofilaments were also extensively investigated for lithium storage [34-37]. A significant reversible capacity has been found, up to 780mAh/g for multiwalled nanotubes (MWNTs) and 1000mAh/g for ball-milled single wall nanotubes (SWNTs), however with some drawbacks. Figure 23.7 presents a galvanostatic insertion-deinsertion curve for nanotubes in 1 mol/L LiPF 6 in EC-DEC (ethylene carbonate-diethylcarbonate) mixture which is quite typical for all this family of materials, including SWNTs, MWNTs, and nanofilaments. In all the cases the irreversible capacity measured after the first discharge-charge cycle is extremely high and varies proportionally to the mesopore volume for the different types of nanotubes investigated [3]. The central canal and/or the mesopores formed by the nanotubes entanglement are surely favoring the SEI formation, allowing an easy access of the voluminous solvated lithium cations to the active surface where they can be transformed into the decomposition products. A similarly high value of Cirr is demonstrated by the mesoporous carbons prepared by a templating procedure in clay minerals [38]. Moreover, during subsequent cycling of the electrodes from nanotubes, an additional passivation layer is formed, demonstrated by a continuous increase of C irr [3, 34]. It means that, in the case of carbons with open mesopores, solvated ions can still penetrate to the interior of the electrode where they can
23.2
Lithium Insertion in Carbon Materials
601
(a)
(b)
Figure 23.6 (a) 002 lattice fringes image on a cross section of a carbonized viscose fiber coated with pyrolytic carbon by CVD from propylene at 900°C. The continuous line shows the separation between a microporous carbon at the bottom (fiber) and the lamellar pyrocarbon at the top. The two insets represent a magnification of selected areas in these two parts. (b) Model showing the disordered nanotexture of the fiber and the lamellar and dense nanotexture of the coating. The coating acts as a barrier which hinders an easy diffusion of species to the core of the fiber. (Adapted from [22]).
be further decomposed. Simultaneously, the reversible capacity also diminishes, and the loss is estimated as ca. 30 % after 10 cycles. For all types of nanotubular materials, a high divergence is observed between the values of insertion and extraction potentials [34, 35], that is commonly called hysteresis. The almost
602
Chapter 23 Electrochemical Energy Storage
3.5 3 2.5
~
:.J en >
2 1.5
lJ.J
0.5
a -0.5
a
500
1000 1500 Capacity (rnA h/g)
2000
Figure 23.7 Representative curve of galvanostatic insertion-deinsertion of lithium into carbon nanotubes. Current load of 20 rnA/g. (Adapted from Ref. [3].)
linear increase of the oxidation potential above ca. 0.5 V vs Li (Fig. 23.7) indicates a progressive evacuation of a wide variety of sites, where lithium has different kinds of interactions with the carbon network. Additionally, the fully lithiated nanotubes represent a well-conducting material due to the charge transfer to carbon, whereas the delithiated materials are significantly less conductive than pristine because of the huge amount of electrically insulating SEI; therefore, the cells' resistance changes rapidly depending on the degree of lithium insertion. The comparison of lithium insertion into rnicroporous and mesoporous carbons, definitively leads to the conclusion that ultrarnicroporous carbons of small ASA are required for limiting the extent ofSEI formation. Although mesoporous carbons look to be attractive on the point of view of their large reversible capacity, the important irreversible capacity and the associated hysteresis preclude their use as active material in Li-ion batteries. Regarding a practical application of carbon nanotubes in these systems, recent results obtained with supercapacitor electrodes demonstrate that they could be useful as a percolating additive (below 10 wt%) to the active carbon material [39]. Moreover, the good mechanical properties of the nanotubular materials allow to preserve a good resiliency of the final anodic material during cycling.
23.2.2.3 Effects of doping on the performance of nanostructured carbons The presence of heteroatoms (boron, nitrogen, oxygen, silicon, phosphorus) substituted to carbon in the graphene layers or occurring as functional groups has been considered for changing the electrochemical performance [40, 41]. The foreign atoms modify the electron donor-acceptor properties of the graphene layers [42], and are consequently expected to affect the interactions
23.2
Lithium Insertion in Carbon Materials
603
during insertion and deinsertion of lithium. The main dopants studied in literature are nitrogen and boron. However, because of the difficulty to determine surely whether the heteroatoms are substituted to carbon, or if they are simply located in interstitial sites, a correct interpretation of the data is impossible. In the case of the nitrogen doped materials, various synthesis procedures have been used in order to study the dependence of the electrochemical characteristics with the nitrogen content. Anodes from nitrogen-containing carbons prepared by chemical vapor disposition (CVD) from acetonitrile, pyridine, and acetyleneammonia mixture demonstrate a reversible capacity ca. 250-300 rnA hi g. The irreversible capacity increases with the nitrogen content whereas the cell capacity shifts to lower voltages compared to pure carbon electrodes [43]. In these CxN materials, a part of nitrogen is substituted for carbon ("lattice nitrogen") and the other is chemically bound to organic molecules ("chemical nitrogen"). The lattice nitrogen acts as a donor, weakening the lithium host bond compared with a pure carbon that explains the lowering of cell voltage. The study of disordered polyacrylonitrile derived carbons revealed a reduction in charge capacity from 380 to 254 rnA hi g but a distinctly faster kinetics of the lithium insertion as the heat treatment temperature increased between 500°C and 1000°C [44]. As the nitrogen content is expected to decrease with the increase of heat-treatment temperature, the reversible capacity seems to be correlated with the amount of nitrogen present in this isotropic carbon material. A similar decrease of reversible capacity is observed when quinoline pitch (QPC) obtained at temperatures from 700°C to 1000°C. However, in the same conditions, the capacity of naphthalene pitch carbons (NPC) decreases more rapidly with temperature, and becomes smaller at 1000°C than their QPC counterpart. The better performance of QPC has been attributed to vacancies left after nitrogen evolution, which create new sites for lithium insertion [45]. The disordered carbonaceous materials prepared by pyrolysis of various nitrogen-containing polymers at 600°C contain graphene nitrogen and conjugated nitrogen which cannot be involved in the irreversible reaction with lithium [46]. The higher the content of nitrogen in the pristine polymeric carbon, the higher the reversible capacity. This increase of capacity mainly results from the graphene nitrogen (N 1s binding energy 398.5 eV). Carbonaceous materials ranging from soft carbons of moderate nitrogen content (up to 2.5 wt%) to typical hard carbons with an excess of nitrogen (up to 6 wt%) have been produced by carbonization of CTP - polyacrylonitrile blends with various ratio of the components [47]. The nitrogen-enriched hard carbons demonstrate a relatively low value of reversible capacity, because of the absence of available nanopores. On the other hand, the irreversible capacity increases with the proportion of nitrogen, especially in the form of pyridinic groups. The lone pair of electrons contribute to the trapping of solvated lithium cations on these groups which act as active sites for the electrolyte decomposition during the first reduction, leading to an enhanced irreversible capacity.
604
Chapter 23 Electrochemical Energy Storage
The slight inconsistency in the results presented above seems to arise from overlapping the contributions of the structural-nanotextural characteristics and the nature of the nitrogen functionality. Taking into account that the N ls X-ray photoelectron spectroscopy (XPS) spectra of the materials must be fitted by four components assigned to pyridinic (N6), pyrroliclpyridonic (N5), quaternary (NQ) , and oxidized nitrogen (NX) functionalities [48], respectively, only a correlation of the electrochemical properties with the proportions of each form could give more reliable information. Nevertheless, for all the nitrogen-enriched carbons, the general tendency from the literature data is an increase ofirreversible capacity with the proportion of nitrogen. If one takes into account that in the best cases the reversible capacity is comparable to that of graphite, this leads to the conclusion that the presence of nitrogen in the carbonaceous materials should be rather precluded for this application. Boron is one of the few elements which are known to be surely substituted to carbon into the carbon lattice. Having one electron less than carbon, boron in substitution can act as an electron acceptor and should be able to facilitate the insertion of electron donors [42, 49], such as lithium. This specific aspect of the nature of boronated carbons has attracted recently a particular attention with respect to the possible performance improvement of carbons used as anode material for rechargeable Li-ion batteries [40, 50]. Boron-substituted carbons, B z C 1- z , with 0 < z < 0.17 were prepared by the CVD method using boron trichloride-benzene mixtures at 900°C, and tested in lithium cells [51]. For z = 0, the reversible capacity is x = 0.65, and it increases with z reaching a value of x = 1.17 (435 rnA hi g) for z = 0.17. Because of the presence of boron, the Fermi level is lowered in a rigid band model, allowing more lithium (which plays as electron donor) to be intercalated. When boron-substituted carbons are produced from acetylene and boron trichloride precursors at 1140°C, the amount of lithium reversibly intercalated increases with the boron content up to a limit close to 13 at% of boron [52]. Lithium intercalation is not so efficient beyond this value ofdoping, suggesting that boron can be substituted in the carbon lattice only up to 13 at%. For a given value of lithium concentration, all the cells with boronated carbons show an increase of voltage of about 0.5 V compared to that in the cell with z = 0 [51, 52] that represents an important drawback. This is interpreted by a strengthening of the chemical bond between the lithium and the boronated carbon host compared with a pure carbon host that contributes to increase the lithium potential in this kind of electrode. Taking into account that the irreversible capacity of the boronated carbons is higher than in the corresponding carbons, it does not seem that this kind of doping is profitable for improving the performance of lithium cells anodes. Almost similar results were found when copyrolysis of QI-free CTP with the borane-pyridine complex (BPC) is used to prepare boronated carbon materials [53]. For a series of cokes calcined at 1000°C, the most striking effect of increasing the boron content is an increase of irreversible capacity ~rr from 0.2 to 0.7. Because a high amount of oxygen is found even in the graphitized boronated carbons, it proves that the incorporated boron and
23.2
Lithium Insertion in Carbon Materials
605
nitrogen (from BPe) induce a strong chemisorption activity ofthe material when exposed to air, which is responsible for the high irreversible capacity in all these materials.
23.2.3 Mechanism of Reversible Li Insertion/Deinsertion in Disordered Carbons Although a very realistic model for lithium intercalation in graphitic materials is proposed since many years [16], the lithium insertion mechanism in disordered carbons remains still not completely documented. The main reason is the complex structure-nanotexture of these carbons which practically makes impossible to propose a unique model applicable for all the materials. Many papers try to interpret the values of reversible capacity in these carbons and to identify the kinds of sites occupied by lithium. An almost linear correlation has been found between the HI C ratio ofvarious polymers pyrolysed below 1000°C and the reversible capacity [54]. The lithium atoms may bind on the hydrogen-terminated edges at the periphery of the basal structural units (BSUs) , causing a change of carbon hybridization from Sp2 to Sp3 during insertion, and from Sp3 to Sp2 during deinsertion [55]. The additional energy which is required for the hybridization change during lithium removal would be the cause of the large potential hysteresis observed with these materials. However, some disordered carbons can maintain capacities higher than 400 rnA hi g, even though their hydrogen content is considerably reduced by a heat treatment [56]. In addition, they demonstrate a low-voltage plateau and little hysteresis [54, 57]. Therefore, several structural-microtextural models have been presented to interpret the reversible insertion behavior of these carbons. From X-ray diffraction and small-angle X-ray scattering measurements, it has been suggested that lithium is adsorbed reversibly onto the internal surfaces of nanopores formed by single-, bi-, and tri-layer groups of graphene sheets arranged like in a "house of cards" [58]. The nanopores size estimated from small-angle X-ray scattering is close to 7 A [59]. In the hypothesis that all the graphene sheets are isolated, the reversible capacity would be twice larger than with graphite [54]. This adsorption on the surface of nanopore walls is similar to the cavity filling by lithium [60]. In the latter case, lithium can as well form clusters in the cavities between the walls, and be intercalated between the graphene layers [61]. Much progress has been done in better determining the location and state of lithium by using ex situ 7Li nuclear magnetic resonance (NMR) at different steps of reduction or oxidation of disordered carbons [62, 63]. This technique supplies also information about the sequence of pores filling by lithium and their empting depending on the potential. However, some doubts may result from the experimental procedure that is used for the realization of ex situ experiments. In particular, the equilibrium composition of the samples can be altered after their extraction from the cell, either by washing or by moisture and oxygen
606
Chapter 23 Electrochemical Energy Storage
from the atmosphere of the glove box. Additionally, each value of cell potential requires the preparation of a new sample, which limits the number of possible experiments. In order to circumvent all these inconvenience, in situ 7Li NMR has been performed on a supple ultrathin plastic carbon-lithium cell at various steps of galvanostatic cycling [57, 64]. The carbon electrode was a composite from ex-viscose carbon fibers coated by a thin layer of pyrolytic carbon, which demonstrate a small irreversible capacity and a reversible capacity 50 % higher than graphite. During the reduction, two lines related with lithium insertion were observed, the minor one at 18 ppm attributed to intercalated lithium (with a charge transfer to carbon), while the position of the most important is downfield shifted during lithium insertion to reach values of ca. 120 ppm, characteristic of quasi-metallic lithium, at full reduction of carbon. After the oxidation step, these two lines disappear and only two other lines remain at 263 and 0 ppm, attributed to metallic lithium from the counter electrode and to the electrolyte-passivation layer, respectively. Figure 23.8 presents a model for the lithium insertionextraction in the carbon fiber, which takes into account the 7Li-NMR data and the structural-nanotextural data provided by high-resolution transmission electron microscopy (HRTEM) observation of the host carbon. Lithium first intercalates in the small intervals between the nanometer-size graphitic type
Insertion
--'\'"
0 (,2
l
~
~
f
~ ~
I
olp
0.2
0.4
0.6
0.8
1.0
-20
1.4
1.6
1.8
,1°
22
J
-40 -60
1.2
I
lJ
i-..."""----""""""'" ,., "'-'
-~
....----
-80 Voltage (V)
Figure 23.16 Voltammetry characteristics of a capacitor built from activated nanotubes at scan rates of 2 and 10mV/ s. Electrolyte 1.4 M TEABF 4 in acetonitrile. (Adapted from Ref. [108].)
23.3.4.3 Use of nanotubes as a support for materials with pseudocapacitance properties Because of their good electrical conductivity and their high mesoporous character, MWNTs are better adapted as a support for electroactive materials with pseudocapacitance properties, e.g., conducting polymers or transition metal oxides. Pseudocapacitance is an intermediate situation where a faradaic charge transfer occurs with a continuous voltage change during charging or discharging, as in a real capacitor [2]. In the case of polypyrrole (PPy) this can be described by the following reaction:
(23.7) where an electron transfer reaction is coupled with the counter ions exchange during the electrochemical oxidation and reduction of the polymer. Capacitance properties of composites from nanotubes with conducting polymers [69, 70, 110-112], e.g., polypyrrole (PPy), polyaniline (PANI), or a polythiophene derivative (PEDOT) have been studied. Polypyrrole (PPy) is the most promising for its high conductivity, stability in the oxidized form and ability to be electrochemically switched between the conducting and the isolating states. PPy can supply a high specific capacitance, because of its high doping level; however, it is valid only for thin films that has a limited practical application. Generally, the thick layers of PPy undergo shrinkage and swelling
620
Chapter 23 Electrochemical Energy Storage
what is the cause of film degradation and conductivity loss. Moreover, bulk PPy supplies very low capacitance values (below 20 Fig) at current loads of 200-400 mAl g. The combination of the two components, using chemical or electrochemical polymerization to get a layer of PPy on the nanotubes, allows all these problems to be overcome and to use efficiently the mesoporous nanotubular network for supplying a perfect three-dimensional volumetric charge distribution [70]. In the case of the electrochemically obtained nanotubeslpolypyrrole (PPy) composites, the values of capacitance reached ca. 170 Fig [69, 70]. Such composite electrodes could be easily cycled at 350 mAl g current load without any performance aggravation over 2000 cycles [100]. Recently, similar composites have been prepared by electrochemical deposition of PPy on well-aligned carbon nanotube electrodes [113, 114]. However, as the values of capacitance presented in this work are given per surface of electrode, it is difficult to estimate the possible applications of this material in a real capacitor. The highest values of capacitance, i.e., 200 Fig with a long durability if a limited voltage range (0.6 V) is selected [115], were obtained from composites containing 80 wt% of PPy deposited chemically on highly entangled catalytic MWNTs. In this case, the PPy layer is more porous, less compact, allowing the diffusion of ions to proceed more easily. Several kinds of polymers were deposited on MWNTs by chemical polymerization and used for supercapacitors. The scanning electron microscopy (SEM) image presented in Fig. 23.17 shows
Figure 23.17 SEM micrograph showing the morphology of a polyaniline/multiwalled nanotubes (PANI/MWNTs) composite material which contains 80 wt % of chemically deposited polyaniline. (Adapted from Ref. [115].)
23.4 General Conclusion and Perspectives
621
the example of the composite with 80 % of deposited PANI which is quite homogeneous. SWNT IPPy nanocomposites have been also investigated in alkaline solution [111]. However, it seems that the nickel foam used for the current collector supplies an additional capacity in this medium. It is also well-known that PPy in alkaline solution degrades quickly, hence, these results have a limited practical application.
23.3.5 Conclusion Contrarily to Li-ion batteries, the supercapacitor application requires highly developed surface area carbons with micropores adapted to the size of the ions involved in the formation of the electric double layer. In this case, the additional presence of mesopores is crucial to fulfill the demand of fast charge propagation with a minimal time constant. It seems that the most suitable would be to increase the amount of mesopores in KOH activated carbons or to increase the microporosity of the essentially mesoporous template carbons. A further improvement of the materials could be a special carbon doping by the incorporation of heteroatoms able to provide useful pseudocapacitance effects. From a careful analysis of the results given by nanotubular materials, it seems that nanotubes can be a perfect component of the electrodes, generally as a support. Carbon nanotubes have a higher rate of electron transfer than conventional carbons due to the curvature of the graphene layers and to their specific open mesoporous texture offering an accessible electrode/electrolyte interface. The novel nanocomposite electrodes, combining the complementary properties of nanotubes and conducting polymers or oxides such as amorphous MnO z [39] are very promising due to the very efficient three-dimensional charges distribution.
23.4
GENERAL CONCLUSION AND PERSPECTIVES
In a nonexhaustive way, this chapter shows that Li-ion batteries and supercapacitors are very important electrical energy storage systems, where the carbon material plays a central role in the performance. Lately, many types of carbons have been investigated more or less empirically in these cells. However, the works performed recently pay a special attention to find correlations with specific parameters of nanostructured carbons, which is rather difficult because of the highly disordered state of these materials.
622
Chapter 23 Electrochemical Energy Storage
From the above presented data, it is obvious that the storage properties are highly influenced by the surface functionality and the porous texture of the materials. For example, we have shown that the irreversible capacity of lithium batteries is mainly related to the (ASA) of the materials. The surface groups play also a significant role in supercapacitors by providing an additional pseudocapacitance, and by reducing considerably the calendar life. The porosity of carbon, i.e., the pore size and the total pore volume, strongly affects the capacity of both the energy storage systems, and some trends show that the performance could be considerably changed by a perfect control of the pore size. Although devices based on carbon electrodes are commercially available, they have to be improved. In the case of lithium anodes, efforts to optimize the surface functionality have to be continued, even for graphite where exfoliation during cycling is a real drawback of graphite-based electrodes. The high values of reversible capacity given by some disordered carbons show promise for future developments. However, it is necessary to find a way to cure the noticeable overvoltage for lithium extraction, which more or less precludes a practical use of these materials. This overpotential, which corresponds to the additional energy requested to extract lithium, could certainly be partly reduced by an optimization of porosity to render lithium withdrawing more easy. In that sense, techniques such as CO 2 adsorption should be more extensively used to investigate the porous texture of these carbons. Regarding the supercapacitors, the real problem is not to increase the capacitance of the materials. Indeed, there are enough commercial carbons giving high values of capacitance. The most fundamental problem for solving is to determine the proportion of the total pore volume that is really used for the process of charge storage. In other words, what are the optimal pore size and shape required for efficient transportation and sorption of ions in a given electrolyte? Answering to this question is ofprime importance, because all the useless volume should be excluded from the material, in order to enhance as much as possible the specific capacity. The nanostructured carbons prepared from templates are certainly a good issue for providing useful information on pore size effects. Another important factor that determines the possible use of supercapacitors is their cycle life. They are supposed to operate for millions of cycles without considerable aging. Actually, this is not the case for supercapacitors operating in organic medium, mainly because of electrolyte decomposition on the surface functionality. Hence, harmful surface groups must be better determined and also the strategies which allow their elimination. Taking into account the underestimated advantages to operate in aqueous electrolyte, it seems also important to look for other applications of carbon materials where the unique combination of electrical conductivity, surface functionality and porous texture may be useful. Such applications as electrochemical hydrogen storage [116, 117], asymmetric supercapacitors [118] open future perspectives where all the information previously collected on other systems will be useful.
References
623
REFERENCES 1. Chau, K.T, Wong, Y.S., and Chan, C.C. (1999). An overview of energy sources for electric vehicles. Ener. Conv. Manag., 40, 1021-39. 2. Frackowiak, E. and Beguin, F. (2001). Carbon materials for the electrochemical storage of energy in capacitors. Carbon, 39, 937-50. 3. Frackowiak, E. and Beguin, F. (2002). Electrochemical storage ofenergy in carbon nanotubes and nanostructured carbons. Carbon, 40, 1775-87. 4. Beguin, F. (2002). Crystallochemistry of intercalation in the crystalline forms of carbon. In Carbon Molecules and Materials (R. Setton, P. Bernier, and S. Lefrant, eds). Taylor and Francis. 5. Wakihara, M. and Yamamoto, o. (eds). (1998). Lithium Ion Batteries - Fundamentals and Peiformance Kodansha Ltd. Wiley-VCH. 6. Conway, B.E. (1999). Electrochemical Supercapacitors - Scientific Fundamentals and Technological Applications. Kluwer Academic/Plenum. 7. Che, G., Lakshmi, B.B., Martin, C.R., and Fisher, E.R. (1999). Metalnanocluster-filled carbon nanotubes: catalytic properties and possible applications in electrochemical energy storage and production. Langmuir, 15, 750-8. 8. Mayer, S.T., Pekala, R.W., and Kaschmitter, J.L. (1993). The aeorocapacitor: an electrochemical double-layer energy-storage device. J. Electrochem. Soc., 140, 446-51. 9. Kyotani, T. (2000). Control of pore structure in carbon. Carbon, 38, 269-86. 10. Ryoo, R., Joo, S.H., Kruk, M., and Jaroniec, M. (2001). Ordered mesoporous carbons. Adv. Mater., 13,677-81. 11. Lee, J., Yoon, S., Hyeon, T., et al. (1999). Synthesis of a new mesoporous carbon and its application to electrochemical double-layer capacitors. Chem. Commun., 2177-8. 12. Fuertes, A.B. (2003). Template synthesis of mesoporous carbons with a controlled particle size. J. Mater. Chem., 13, 3085-8. 13. Marsh, H., Yan, D.S., O'Grady, T.M., and Wennerberg, A.N. (1984). Formation of active carbons from cokes using potassium hydroxide. Carbon, 22, 603-11. 14. Lahaye,J., Dentzer,J., Soulard, P., and Ehrburger, P. (1991). Carbon gasification: the active site concept. In Fundamental Issues of Control of Carbon Gasification Reactivity O. Lahaye and P. Ehrburger, eds). Academic Publishers, pp. 143-58. 15. Sawai, K., Iwakoshi, Y., and Ohzuku T. (1994). Carbon materials for lithium-ion (shuttlecock) cells. Solid State Ionics, 69, 273- 83. 16. Guerard, D. and Herold, A. (1975). Intercalation of lithium into graphite and other carbons. Carbon, 13, 337-45. 17. Winter, M., Besenhard, J.O., Spahr, M.E., and Novak, P. (1998). Insertion electrode materials for rechargeable lithium batteries. Adv. Mater., 10, 725-63. 18. Naji, A., Ghanbaja, J., Humbert, B., Willmann, P., and Billaud, D. (1996). Electroreduction of graphite in LiClO 4 -ethylene carbonate electrolyte. Characterization of the passivating layer by transmission electron microscopy and Fourier-transform infrared spectroscopy. J. Power Sourc., 63, 33-9. 19. Aurbach, D., Zaban, A., Ein-Eli, Y., et al. (1997). Recent studies on the correlation between surface chemistry, morphology, three-dimensional structures and
62 4
20.
21.
22.
23.
24. 25.
26.
27.
28.
29.
30. 31.
32. 33.
34. 35.
Chapter 23 Electrochemical Energy Storage
performance of Li and Li-C intercalation anodes in several important electrolyte systems.]. Power Sourc., 68, 91-8. Kanamura, K., Tamura, H., Shiraishi, S., and Takehara, Z. (1995). Morphology and chemical compositions of surface films of lithium deposited on a Ni substrate in nonaqueous electrolytes.]. Electroanal. Chem., 394, 49-62. Besenhard, J.O., Winter, M., Yang, J., and Biberracher, W. (1995). Filming mechanism of lithium-carbon anodes in organic and inorganic electrolytes.]. Power Sourc., 54,228-31. Beguin, F., Chevallier, F., Vix, C., et al. (2004). A better understanding of the irreversible lithium insertion mechanisms in disordered carbons.]. Phys. Chem. Solids, 65, 211-17. Larcher, D., Mudalige, C., Gharghouri, M., and Dahn,J.R. (1999). Electrochemical insertion of Li and irreversibility in disordered carbons prepared from oxygen and sulfur-containing pitches. Electrochim. Acta, 44, 4069-72. Xing, W. and Dahn, J.R. (1997). Study of irreversible capacities for Li insertion in hard and graphitic carbons.]. Electrochem. Soc., 144, 1195-201. Guerin, K., Fevrier-Bouvier, A., Flandrois, S., et al. (2000). On the irreversible capacities of disordered carbons in lithium-ion rechargeable batteries. Electrochim. Acta, 45, 1607-15. Guerin, K., Menetrier, M., Fevrier-Bouvier, A., et al. (2000). 7Li NMR study of a hard carbon for lithium-ion rechargeable batteries. Solid State Ionics, 127, 187-98. Winter, M., Novak, P., and Monnier, A. (1998). Graphites for lithium-ion cells: the correlation of the first-cycle charge loss with the Brunauer-Emmett-Teller surface area.]. Electrochem. Soc., 145, 428-36. Fong, R., Von Sacken, U., and Dahn, J.R. (1990). Studies oflithium intercalation into carbons using nonaqueous electrochemical cells. J. Electrochem. Soc., 137, 2009-13. Simon, B., Flandrois, S., Fevrier-Bouvier, A., and Biensan, P. (1998). Hexagonal vs rhombohedral graphite: the effect of crystal structure on electrochemical intercalation of lithium ions. Mol. Cryst. Liq. Cryst., 310, 333-40. Laine, N.R., Vastola, F.J., and Walker, P., Jr (1963). The importance of active surface-area in the carbon-oxygen reaction.]. Phys. Chem., 67, 2030-4. Beguin, F., Chevallier, F., Letellier, M., et al. (2006). Mechanism of reversible and irreversible insertion in nanostructured carbons used in lithium-ion batteries. In New Carbon Based Materials for Electrochemical Energy Storage Systems, NATO Science Series, Mathematics, Physics and Chemistry, Springer, Dordrecht: The Netherlands, 229, 231-43. Kuribayashi, I., Yokoyama, M., and Yamashita, M. (1995). Battery characteristics with various carbonaceous materials.]. Power Sourc., 54, 1-5. Yoon, S., Kim, H., and Oh, S.M. (2001). Surface modification of graphite by coke coating for reduction of initial irreversible capacity in lithium secondary batteries.]. Power Sourc., 94, 68-73. Frackowiak, E., Gautier, S., Gaucher, H., et al. (1999). Electrochemical storage of lithium in multiwalled carbon nanotubes. Carbon, 37, 61-9. Gao, B., Bower, C., Lorentzen, J.D., et al. (2000). Enhanced saturation lithium composition in ball-milled single-walled carbon nanotubes. Chem. Phys. Lett., 327,69-75.
References
625
36. Wu, G.T., Wang, C.S., Zhang, X.B., et al. (1999). Structure and lithium insertion properties of carbon nanotubes. J. Electrochem. Soc., 146, 1696-701. 37. Claye, A.S., Fischer, J.E., Huffman, C.B., et al. (2000). Solid-state electrochemistry of the Li single wall carbon nanotube system. J. Electrochem. Soc., 147, 2845-52. 38. Duclaux, L., Frackowiak, E., and Beguin, F. (1999). Novel carbons from nanocomposites for high lithium storage. J. Power Sourc., 81-82, 323-7. 39. Raymundo-Pinero, E., Khomenko, V., Frackowiak, E., and Beguin, F. (2005). Performance ofmanganese oxide/carbon nanotubes composites as electrode materials for electrochemical capacitors. J. Electrochem. Soc., 152, A229-A235. 40. Flandrois, S. and Simon, B. (1999). Carbon materials for lithium-ion rechargeable batteries. Carbon, 37, 165-80. 41. Wu, Y.P., Rahm, E., and Holze, R. (2002). Effects ofheteroatoms on electrochemical performance of electrode materials for lithium ion batteries. Electrochim. Acta, 47, 3491-507. 42. Marchand, A. (1971). Electronic properties of doped carbons. In Chemistry and Physic of Carbon, Vol. 7 (P.L. Walker,Jr, ed.). Marcel Dekker, pp. 155-91. 43. Weydanz, W.J., Way, B.M., van Buuren, T., and Dahn, J.R. (1994). Behavior of nitrogen-substituted carbon (NzC 1- z ) in Li/Li(Nz C 1_z )6 cells. J. Electrochem. Soc., 141, 900-7. 44. Jung, Y., Suh, M.C., Lee, H., et al. (1997). Electrochemical insertion of lithium into polyacrylonitrile-based disordered carbons. J. Electrochem. Soc., 114, 4279-84. 45. Mochida, I., Ku, C., Yoon, S., and Korai, Y. (1999). Anodic performances of anisotropic carbon derived from isotropic quinoline pitch. Carbon, 37, 323-7. 46. Wu, Y., Fang, S., and Jiang, Y. (1999). Effects of nitrogen on the carbon anode of a lithium secondary battery. SoUd State Ionics, 120, 117-23. 47. Machnikowski, J., Grzyb, B., Weber, J.V., et al. (2004). Structural and electrochemical characterization of nitrogen enriched carbons produced by the copyrolysis of coal-tar pitch with polyacrylonitrile. Electrochim. Acta, 49, 423-32. 48. Kapteijn, F., Moulijn, J.A., Matzner, S., and Boehm, H.-P. (1999). The development of nitrogen functionality in model chars during gasification in CO 2 and 02' Carbon, 37, 1143-50. 49. Flandrois, S., Ottaviani, B., Derre, A., and Tressaud, A. (1996). Boron substituted carbons and their intercalation compounds. J. Phys. Chern. Solids, 57, 741-4. 50. Endo, M., Kim, C., Nishimura, K., et al. (2000). Recent development of carbon materials for Li ion batteries. Carbon, 38, 183-97. 51. Way, B.M. and Dahn, J.R. (1994). The effect of boron substitution in carbon on the intercalation of lithium in Lix (B z C 1- z )6' J. Electrochem. Soc., 141, 907-12. 52. Shirasaki, T., Derre, A., Guerin, K., and Flandrois, S. (1999). Chemical and electrochemical intercalation of lithium into boronated carbons. Carbon, 37, 1961-4. 53. Machnikowski, J., Frackowiak, E., Kierzek, K., et al. (2004). Lithium insertion into boron containing carbons prepared by co-pyrolysis of coal-tar pitch and borane-pyridine complex. J. Phys. Chern. Solids, 65, 153-8. 54. Zheng, T., Liu, Y., Fuller, E.W., et al. (1995). Lithium insertion in high capacity carbonaceous materials. J. Electrochem. Soc., 142, 2581-90. 55. Zheng, T., Zhong, Q., and Dahn, J.R. (1996). Hysteresis during insertion in hydrogen-containing carbons. J. Electrochem. Soc., 143, 2137-45. 56. Liu, Y., Xue, J.S., and Dahn, J.R. (1996). Mechanism of lithium insertion in hard carbons prepared by pyrolysis of epoxy resins. Carbon, 34, 193-200.
626
Chapter 23 Electrochemical Energy Storage
57. Chevallier, F., Letellier, M., Morcrette, M., et al. (2004). In situ 7Li Nuclear Magnetic Resonance observation of reversible lithium insertion into disordered carbons. Electrochem. Solid State Lett., 6, A225-8. 58. Zheng, T., Xue, J.S., and Dahn J.R. (1996). Lithium insertion in hydrogencontaining carbonaceous materials. Chem. Mater., 8, 389-93. 59. Gibaud, A., Xue, J.S., and Dahn, J.R. (1996). A small angle X ray scattering study of carbons made from pyrolyzed sugar. Carbon, 34, 499-503. 60. Tokumitsu, K., Mabuchi, A., Fujimoto, H., and Kasuh, T. (1996). Electrochemical insertion of lithium into carbons synthesized from condensed aromatics. ]. Electrochem. Soc., 143, 2235-9. 61. Mabuchi, A., Fujimoto, H., Tokumitsu, K., and Kasuh, T. (1995). Chargedischarge characteristics of the mesocarbon microbeads heat-treated at different temperatures.]. Electrochem. Soc., 142, 1041-6. 62. Gautier, S., Leroux, F., Frackowiak, E., et al. (2003). Influence of the pyrolysis conditions on the nature of lithium inserted in hard carbons.]. Phys. Chem. A, 105, 5794-800. 63. Tatsumi, K., Akai, T., Imamura, T., et al. (1996). 7Li-Nuclear magnetic resonance observation of lithium insertion into mesocarbon microbeads.]. Electrochem. Soc., 143, 1923-30. 64. Letellier, M., Chevallier, F., Clinard, C., et al. (2003). The first in situ 7Li nuclear magnetic resonance study of lithium insertion in hard-carbon anode materials for Li-ion batteries.]. Chem. Phys., 118,6038-45. 65. Conway, B.E. (1991). Transition from "supercapacitor" to "battery" behavior in electrochemical energy storage.]. Electrochem. Soc., 138, 1539-48. 66. Toupin, M., Brousse, T., and Belanger, D. (2002). Influence of microtexture on the charge storage properties of chemically synthesized manganese dioxide. Chem. Mater., 14, 3946-52. 67. Wu, N.L. (2002). Nanocrystalline oxide supercapacitors. Mater. Chem. Phys., 75, 6-11. 68. Miller, J.M., Dunn, B., Tran, T.D., and Pekala, R.W. (1999). Morphology and electrochemistry of ruthenium/carboniaerogel nanostructures. Langmuir, 15, 799-806. 69. Frackowiak, E., Jurewicz, K., Delpeux, S., et al. (2002). Synergy of components in supercapacitors based on nanotube/polypyrrole composites. Mol. Cryst. Liq. Cryst., 387, 73-8. 70. Jurewicz, K., Delpeux, S., Bertagna, V., et al. (2001). Supercapacitors from nanotubes/polypyrrole composites. Chem. Phys. Lett., 347, 36-40. 71. Laforgue, A., Simon, P., Sarrazin, Ch., and Fauvarque, J-F. (1999). Polythiophene-based supercapacitors.]. Power Sourc., 80, 142-8. 72. Arbizzani, C., Mastragostino, M., and Soavi, F. (2001). New trends in electrochemical supercapacitors.]. Power Sourc., 100, 164-70. 73. Mastragostino, M., Arbizzani, C., and Soavi, F. (2002). Conducting polymers as electrode materials in supercapacitors. Solid State Ionics, 148, 493-8. 74. Jurewicz, K., Babel, K., Ziolkowski, A., et al. (2002). Ammoxidation of brown coals for supercapacitors. Fuel Process. Tech., 77-78, 191-8. 75. Jurewicz, K., Babel, K., Ziolkowski, A., and Wachowska, H. (2003). Ammoxidation of active carbons for improvement of supercapacitor characteristics. Electrochim. Acta, 48, 1491-8.
References
627
76. Qu, D. and Shi, H. (1998). Studies of activated carbons used in double-layer capacitors.]. Power Sourc., 74, 99-107. 77. Gambly,]., Taberna, P.L., Simon, P., et al. (2001). Studies and characterization of various activated carbons used for carbon/carbon supercapacitors]. Power Sourc., 101, 109-16. 78. Shi, H. (1996). Activated carbons and double layer capacitance. Electrochim. Acta, 41, 1633-9. 79. Salitra, G., Soffer, A., Eliad, L., et al. (2000). Carbon electrodes for double-layer capacitors. I. Relations between ions and pore.]. Electrochem. Soc., 147,2486-93. 80. Guo, Y., Qi, J., Jiang, Y., et al. (2003). Performance of electrical double layer capacitors with porous carbons derived from rice husk. Mater. Chem. Phys., 80, 704-9. 81. Okamura, M. (2000). US Patent 6064562 and JP11067608 forJEOL Ltd. 82. Lozano-Castello, D., Lillo-Rodenas, M.A., Cazorla-Amoros, D., and LinaresSolano, A. (2001). Preparation of activated carbons from Spanish anthracite I. Activation by KOH. Carbon, 39, 741-9. 83. Lozano-Castello, D., Cazorla-Amoros, D., Linares-Solano, A., et al. (2003). Influence of pore structure and surface chemistry on electric double layer capacitance in non-aqueous electrolyte. Carbon, 41, 1765-75. 84. Kierzek, K., Frackowiak, E., Lota, G., et al. (2004). Electrochemical capacitors based on highly porous carbons prepared by KOH activation. Electrochim. Acta, 49,515-23. 85. Raymundo-Pinero, E., Cazorla-Amoros, D., Linares-Solano, A., et al. (2002). High surface area carbon nanotubes prepared by chemical activation. Carbon, 40, 1614-7. 86. Vix-Guterl, C., Saadallah, S., Jurewicz, K., et al. (2004). Supercapacitor electrodes from new ordered porous carbon materials obtained by a templating procedure. Mater. Sci. Eng. B, 108, 148-55. 87. Han, S., Lee, K.T., Oh, S.M., and Hyeon, T. (2003). The effect of silica template structure on the pore structure of mesoporous carbons. Carbon, 41, 1049-56. 88. Jurewicz, K., Vix, C., Frackowiak, E., et al. (2004). Capacitance properties of ordered porous carbon materials prepared by a templating procedure. J. Phys. Chern. Solids, 65, 287-93. 89. Yoon, S., Lee,J., Hyeon, T., and Oh, S.M. (2000). Electric double-layer capacitor performance of a new mesoporous carbon. J. Electrochem. Soc., 147, 2507-12. 90. Vix-Guterl, C., Frackowiak, E., Jurewicz, K., et al. (2005). Electrochemical energy storage in ordered porous carbon materials. Carbon, 43, 1293-302. 91. Endo, M., Kim, Y.J., Ohta, H., et al. (2002). Morphology and organic EDLC applications ofchemically activated AR-resin-based carbons. Carbon, 40, 2613-26. 92. Conway, B.E., Verall, R.E., and Desnoyers,J.E. (1966). Trans. Faraday Soc., 62, 2738. 93. Robinson, R.A. and Stokes R.H. (1965). Electrolyte Solutions. Butterworths. 94. Pell, W.G., Conway, B.E., and Marincic, N. (2000). Analysis of non-uniform charge/discharge and rate effects in porous carbon capacitors containing suboptimal electrolyte concentrations. J. Electroanal. Chern., 491, 9-21. 95. Niu, C., Sichel, E.K., Hoch, R., et al (1997). High power electrochemical capacitors based on carbon nanotube electrodes. Appl. Phys. Lett., 70, 1480-2.
628
Chapter 23 Electrochemical Energy Storage
96. Delpeux, S., Szostak, K., Frackowiak, E., et al. (2002). High yield of pure multiwalled carbon nanotubes from the catalytic decomposition of acetylene on in-situ formed cobalt nanoparticles. J. Nanosci. Nanotech., 2, 481-4. 97. Ma, R.Z., Liang, J., Wei, B.Q., et al. (1999). Study of electrochemical capacitors utilizing carbon nanotube electrodes. J. Power Sourc., 84, 126-9. 98. Frackowiak, E., Metenier, K., Bertagna, V., and Beguin F. (2000). Supercapacitor electrodes from multiwalled carbon nanotubes. Appl. Phys. Lett., 77, 2421-3. 99. Frackowiak, E., Jurewicz, K., Delpeux, S., and Beguin, F. (2001). Nanotubular materials for supercapacitors. J. Power Sourc., 97-98, 822-5. 100. Frackowiak, E., Jurewicz, K., Szostak, K., Delpeux, S., and Beguin, F. (2002). Nanotubular materials as electrodes for supercapacitors. Fuel Process. Tech., 77-78, 213-9. 101. Chen, J.H., Li, W.Z., Wang, D.Z., et al. (2002). Electrochemical characterization of carbon nanotubes as electrode in electrochemical double-layer capacitors. Camon, 40,1193-7. 102. Zhang, B., Liang, J., Xu, C.L., et al. (2001). Electric double-layer capacitors using carbon nanotube electrodes. Mater. Lett., 51, 539-42. 103. An, K.H., Kim, W.S., Park, Y.S., et al. (2001). Supercapacitors using single-walled carbon nanotube electrodes. Adv. Mater., 13, 497-500. 104. Shiraishi, S., Kurihara, H., Okabe, K., et al. (2002). Electric double layer capacitance of highly pure single-walled carbon nanotubes (HiPco™ BuckytubesTM) in propylene carbonate electrolytes. Electrochem. Commun., 4, 593-8. 105. Frackowiak, E. (2004). Carbon nanotubes for storage of energy: supercapacitors. Encyclopedia of Nanoscience and Nanotechnology. Marcel Dekker, pp. 537-46. 106. Barisci, J.N., Wallace, G.G., and Baughman, R.H. (2000). Electrochemical quartz crystal microbalance studies of single-wall carbon nanotubes in aqueous and nonaqueous solutions Electrochim. Acta., 46, 509-17. 107. Endo, M., Kim, Y.J., Osawa, K., et al. (2003). High capacitance EDLC using a carbon material obtained by carbonization of PVDC: the effect of the crystallite size of the pristine PVDC. Electrochem. Solid State Lett., 6, A23. 108. Frackowiak, E., Delpeux, S., Jurewicz, K., et al. (2002). Enhanced capacitance of carbon nanotubes through chemical activation. Chem. Phys. Lett., 361, 35-41. 109. Jiang, Q., Qu, M.Z., Zhou, G.M., et al. (2002). A study of activated carbon nanotubes as electrochemical super capacitors electrode materials. Mater. Lett., 57, 988-91. 110. Lota, K., Khomenko, V. and Frackowiak, E. (2004). Capacitance properties of poly(3,4-ethylenedioxythiophene)/carbon nanotubes composites J. Phys. Chem. Solids, 65, 295-301. 111. An, K.H., Jeon, K.K., Heo, J.K., et al. (2002). High-capacitance supercapacitor using a nanocomposite electrode of single-walled carbon nanotube and polypyrrole. J. Electrochem. Soc., 149, Al058-62. 112. Fan, J., Wan, M., Zhu, D., et al. (1999). Synthesis, characterizations, and physical properties of carbon nanotubes coated by conducting polypyrrole.J. Appl. Polymer Sci., 74, 2605-10. 113. Chen, J.H., Huang, Z.P., Wang, D.Z., et al. (2001). Electrochemical synthesis of polypyrrolelcarbon nanotube nanoscale composites using well-aligned carbon nanotube arrays. Appl. Phys. A Mater. Sci. Process, 73, 129-31.
References
62 9
114. Chen, J.H., Huang, Z.P., Wang, D.Z., et al. (2002). Electrochemical synthesis of polypyrrole films over each of well-aligned carbon nanotubes. Synth. Met., 125, 289-94. 115. Khomenko, V., Frackowiak, E., Szostak, K., and Beguin, F. (2005). Determination of specific capacitance of conducting polymers/nanotubes composite electrodes using different cell configurations. Electrochim. Acta, 50, 2499-506. 116. Jurewicz, K., Frackowiak, E., and Beguin, F. (2001). Enhancement of reversible hydrogen capacity into activated carbon through water electrolysis. Electrochem. Solid State Lett., 4, A27-9. 117. Jurewicz, K., Frackowiak, E., and Beguin, F. (2004). Towards the mechanism of electrochemical hydrogen storage in nanostructured carbon materials. Appl. Phys. A, 78, 981-7. 118. Brousse, T., Toupin, M., and Belanger. (2004). A hybrid activated carbonmanganese dioxide capacitor using a mild aqueous electrolyte. J. Electrochem. Soc., 151~ 614-22.
ADSORPTION OF INORGANIC SPECIES FROM
AQUEOUS SOLUTIONS Pierre Le Cloirec and Catherine Faur-Brasquet Ecole des Mines de Nantes, Nantes cedex, France
Contents 24.1 Introduction 24.2 Metal Ion Removal 24.3 Anion and Cation Removal 24.4 Reaction Between Activated Carbon and Oxidants 24.5 Catalytic Reactions with Modified Activated Carbon 24.6 Conclusions and Trends References
24.1 INTRODUCTION Water and wastewater can be considered as complex mixtures of suspended solids, colloids, and dissolved organic or inorganic pollutants due to natural discharges or human activities. The contaminant levels are quite low in drinking water sources compared to pollutant concentrations found in industrial wastewater. However, to obtain clean water, several physicochemical or biological processes are available and commonly carried out, such as sedimentation, coagulation, flocculation, filtration, adsorption, oxidation, and free or fixed microorganisms [1]. To control and limit the impact of inorganic species on human health and the environment, treatment processes have to be defined and proposed. The methods for the removal of cations or anions from water are precipitation, membrane processes (nanofiltration or reverse osmosis), oxidation, biotreatments, ion exchange, and adsorption [2]. Activated carbon in Adsorption by Carbons ISBN: 978-0-08-044464-2
© 2008 Elsevier Ltd. All rights reserved.
631
Chapter 24 Adsorption of Inorganic Species from Aqueous Solutions
the form of powder, grains and, more recently, fibers (cloth or felt) [3] is a universal adsorbent and, in particular, some interactions occur with inorganic species present in water. The purpose of this chapter is to address the mechanisms of these interactions and to describe the processes for the removal of inorganic materials onto activated carbon. First, metal ion removal will be considered for the example of virgin activated carbon or precoated with organic matter. In this case, the regeneration of saturated materials will be described. The second section will be dedicated to cation and anion removal by adsorption, ion exchange, or biological activated carbon (BAC). The activated carbon can react with residual oxidant concentrations in water. In the third part, oxidation-reduction reactions in the presence or absence of catalysts will be presented. Finally, the interactions of porous carbon with metal or metal oxides to give catalytic or photocatalytic reactions will be approached in terms of the production of such material and its use in specific organic treatments.
24.2 METAL ION REMOVAL Environmental pollution due to the significant release of heavy metals by several industries is of major concern because of their toxicity and the threat to human life and to the environment, especially when tolerance levels are exceeded [4]. Besides ion exchange or reverse osmosis, adsorption onto activated carbon is one of the tertiary processes that may be used to remove low concentrations of metals from waste streams or drinking water. Adsorption has been shown to be economically favorable, compared with ion exchange, and technically easy, compared with reverse osmosis [5]. However, although adsorption by activated carbon is a common treatment for organic compound removal, it has been rarely used for metal ion elimination in an industrial setting, despite the fact that its performance has been demonstrated by numerous researchers [6]. Generally speaking, metal ion adsorption onto activated carbon may be studied in terms of distinct but interrelated phenomena: adsorption, surface precipitation, complexation, and ion exchange [7]. The maximum extent of adsorption depends on the nature of the metal ion, given by its speciation in aqueous solution, and on the activated carbon surface chemistry (see Chapter 13). This section aims to give an overview of adsorption processes of metal ions by activated carbon. Three cases are detailed: (1) the adsorption of metal ions onto virgin activated carbon; adsorption capacities are given in static and dynamic reactors, and the influence of various operating conditions is shown; (2) the adsorption of metal ions onto activated carbon preloaded with organic matter; (3) the saturation of activated carbon by organic matter and metal hydroxides after its use in wastewater treatment. The influence of metal hydroxides on activated carbon regeneration is demonstrated.
24.2
Metal Ion Removal
24.2.1
633
Adsorption of Metal Ions by Virgin Activated Carbon
24.2.1.1
Adsorption mechanism
A variety of distinct but interrelated phenomena may be involved in the adsorption process of metal ions onto activated carbons: adsorption (physical adsorption or chemisorption), surface precipitation, ion exchange, and surface complexation. The metal sorption is often not the result of one mechanism but of several reactions. The mechanisms involved and their degree of importance seem to depend on the materials and the operating conditions used. • Different studies have shown, during metal ion removal by activated carbon, a decrease in final pH as metal ion concentration is improved [8]. This fact may be due to a release of H 3 0+ ions and may indicate an adsorption mechanism by ion exchange, expressed by the following reaction where Mm+ is the metal ion and S-OH is a surface functional group of activated carbon [9]:
Ion exchange consists of the replacement of one adsorbed, readily exchangeable ion by another. The involvement ofoxygen surface groups in the sorption mechanism by ion exchange was confirmed by an improvement in adsorption onto oxidized adsorbents [10]. Carboxylic surface groups have been shown to be especially involved in the adsorption process [8], but this phenomenon may also occur with other ions such as [11] Ca2+, K+, and Na2+. • The formation of suiface complexes may also happen, by assuming an amphoteric behavior of activated carbon, according to the following reactions [11]:
for bidentate complexes:
2
OH + M 2+ ~== S - OM+ + H+ S - OH+M 2+ ~ (== S - 02M+2H+)
== S -
for monodentate complexes:
==
• In the presence of a high cation· to sorbent ratio, and at high sorbate concentrations, surface sites become saturated and surface complexation may be replaced by suiface precipitation, which involves the formation of a new solid or gel metal hydroxide at the surface [12]. • An adsorption process may also occur by surface reaction between the cation M 2+ and the negatively charged surface of the activated carbon, without exchange of ions or electrons. The variety of mechanisms that may be involved in the sorption process of metal ions onto activated carbon induces a great number of factors that control the adsorption: the surface oxygen complex content, the pH of point of zero charge, the pore texture of carbon, the solution pH and its ionic strength, the adsorption temperature, the nature of the metal ion given by its speciation diagram, its solubility, and its size in adsorption conditions. The influence of these various conditions is detailed in Section 24.2.1.4.
Chapter 24 Adsorption of Inorganic Species from Aqueous Solutions
24.2.1.2
Adsorption capacities in a batch reactor
Monocomponent adsorption capacities
Although activated carbon is rarely used for metal ion removal in an industrial setting, various works have shown its ability to remove these inorganic pollutants from water. Table 24.1 summarizes some adsorption capacities of activated carbon obtained for different metal ions in a static reactor. Depending on the operating conditions and the activated carbon and metal ion characteristics [6], the adsorption capacity ranges between 2 and 200 mg/g. These values are higher than those obtained with clay, and of the same order as those of sorbents like biomass or peat, the maximum adsorption capacities (up to 1000 mg/g for H~+ or Pb 2+) being reached with chitosan or lignin [13]. Table
24.1 Adsorption capacities of different metal ions onto activated carbon in a static reactor
Co (mg/L) (g/L) pH Q (mg/g) Reference wAC
1 0.2 7 2.6 [14]
Co - initial concentration, na - not available.
10-40 2 5 5-200
100 1 7
[8] WAC -
8
10-40 2 5 5-30
10-40 2 5 2-60
[15]
[8]
[8]
20-1000 10
20-1000 10
na
na
3-11 [16]
20-70 [16]
activated carbon weight, Q (mg/g) - adsorption capacity,
Multicomponent adsorption capacities
The competitive adsorption of metal ions is dependent on both the metal ions and the adsorbent, its magnitude being related to the adsorption mechanism. In the case of competition between metal ions for the same adsorption sites, it has been shown that the favored metal ion is that which presents the faster adsorption kinetics on the same activated carbon in a monocomponent solution. This is the case for the adsorption of Cu(II) and Pb(II) onto activated carbon cloths [17]. When metal ions present in solution do not interact with the same adsorption sites, the removal of both ions is not affected compared with monocomponent adsorption. For example, in a study performed with different activated carbons, nickel removal was not affected by the presence of cadmium, because the sites that interact with nickel do not strongly interact with cadmium [18]. Finally, due to the strong relationship that exists between the metal ion adsorption mechanism and pH (see Section 24.2.1.4), it as been demonstrated that competitive adsorption is also influenced by pH [19]. 24.2.1.3
Adsorption capacities in a column
The adsorption performance of activated carbon was confirmed in an activated carbon column for various metal ions like Cd(II), Pb(II), Hg(II), and
24.2
Metal Ion Removal
Table
24.2
635
Adsorption properties of metal ions by activated carbon in a
column [6]
Co (mg/L) WAC (g) pH U (m/h) Adsorption
112 40 7 4.9 BV = 2.6
52 30 2.5 4.9 BV = 400
50 50 4 3.7 BV= 20
2 na 8 4.6 60% removal
43 120 7 na V=2L
Co - initial concentration, WAC - activated carbon weight, U - hydraulic loading rate, BV - bed volumes treated at breakthrough, V - volume treated at breakthrough, na not available.
Zn(II) as presented in Table 24.2 [6], or Cr(VI), more than 99% of which was removed from industrial electroplating wastewater [20]. The modeling of the experimental breakthrough of lead (II) onto activated carbon fibers in a fixed bed, using axial dispersion and diffusion equations solved by the orthogonal collocation method, demonstrated that the intraparticle and external mass transfer is not the rate-controlling step, due to the short diffusion path for the adsorbate in activated carbon fibers [21].
24.2.1.4 Influence of operating conditions The stronger effect on the adsorption capacities ofmetal ions onto activated carbon is due to the pH contribution, which can be attributed to interactions between ions in solution and complexes formed at the adsorbent surface. The fact that a metal ion in aqueous solution can form different species whose presence depends on the solution pH is well documented [22]. Furthermore, the surface charge of an activated carbon surface also depends on solution pH: the surface charge is positively charged when the solution pH is below the pH of point of zero charge (pHpzc ) and negatively charged when the pH is above pHpzc (see Chapter 13). Different surveys show that a pH increase leads to an improvement in metal ion adsorption. Figure 24.1 presents copper adsorption onto an activated carbon cloth at different concentrations [8]. This result shows that the removal of heavy metals increases from about 10 to 90 % in a narrow pH range, known as the "pH adsorption edge," which is shifted to more alkaline regions as the molar metal concentration increases [19]. The coinfluence of metal ion removal by precipitation is also presented in Fig. 24.1. Table 24.3 summarizes the values of pH adsorption edges for different metals and given operating conditions. The adsorption increase of Cu(II) in a GAC column for a pH increase from 2 to 6 was explained by a shift of the equation (Mm+ + SOH -+ SOM(m-l)+ + H+) from left to right, which results in the production of more surface complexes SOM (m-l)+ or a higher removal of metal ions [9].
636
Chapter 24 Adsorption of Inorganic Species from Aqueous Solutions
100 90 80 70 (ij
60
> 0 E
50
0>R.
40
~
30 20 10
-+---------J'--I--------....--4----l
---+-
% adsorbed - 20 ppm
-+--------J~------4---J-----l ----0---
%
---+-
%
-+-----~---------I'---7'-----l _
-0- -
%
precipitated - 20 ppm adsorbed - 40 ppm precipitated - 40 ppm
0 2
0
4
6
8
10
12
Initial pH
Figure 24.1 Influence of pH on copper adsorption onto an activated carbon cloth [8].
Table 24.3 pH adsorption edges for the adsorption of different metal ions onto activated carbon (AC)
Zn(II) Cu(II) Pb(II) Ni(II)
100
1
40 40 40
2
2-7 2-6
2 2
4-9
3-4
[15] [8,9] [8] [8]
Because of a mechanism by ion exchange or surface complexation, the influence of other ions in solution cannot be neglected. An increase in the ionic strength of the solution reduces the electrostatic interaction, either repulsive or attractive, between the activated carbon surface and the metal ions due to a screening effect of the electrolyte. When the initial electrostatic interaction is repulsive (respectively attractive), an increase in ionic strength induces an increase (respectively a decrease) in adsorption [23]. The type of background electrolyte (NaN0 3 or NaCI0 4) does not significantly affect metal removal [18]. Among other factors that may influence the removal ofmetal ions by activated carbon, different surveys have pointed out the combined effect of initial metal concentration and activated carbon dosage with a decrease in adsorption as the metal/carbon ratio increases [24,25], or of temperature [26].
24.2.1.5 Modeling by surface complexation models Surface complexation models (SCM) are surface chemical equilibrium models that originate from work with metal oxides. The basic premise of SCM
24.2
637
Metal Ion Removal
is that adsorption of ions onto hydrous solids is analogous to the formation of soluble complexes, according to the following single metal-surface complexation reaction:
Various types of SCM have been assessed; namely, the diffuse-layer model (DLM) [27], the constant-capacitance model [28], the Stern model [29], and the triple-layer model (TLM) [30]. They differ in complexity from the simplest, DLM, which has four adjustable parameters, to the most complex, TLM, which includes seven adjustable parameters. The number ofparameters is dependent on the hypothesis relative to the model. In various researches, the DLM is selected because of its simplicity and its applicability to various solution conditions [31]. It takes into account ionic strength effects on protolysis equilibria through the Gouy-Chapman-Stern-Grahame charge-potential relationship:
where (J"o is the surface charge (C/m 2 ), 8 is the dielectric constant of water (78.5 at 25°C), 8 0 is the permittivity of free space (8.854 x 10- 12 C/V1m), Z is the valence of the electrolyte, and I is the ionic strength (moI/L). The DLM model needs to determine four parameters to enable surface complexation computation using the FITEQL program from adsorption vs pH curves [32]. These parameters are • both surface acidity constants of the ionization reaction of surface sites defined by the following equations:
== SOH2+ ~== SOH+H+ == SOH ~== SO- +H+ • the total number of acidic surface sites assessed by the Boehm method (see Chapter 13); • the specific surface area determined by nitrogen adsorption at 77 K. Because of a sorption mechanism of metal ions onto activated carbon that involves surface complexation, the DLM model was successfully applied to describe adsorption of metal ions onto this adsorbent. For example, the adsorption of Cd2+ and Zn2+ onto activated carbon in the form of powder or granules was modeled [19]. The modeling of the adsorption of Cu2+, Ni2+, and Pb 2+ onto activated carbon cloths allowed complexation constants to be calculated:
.
K~ =SOM+
{== SOM+}{H+}
= {== SOH}{M2+} exp
(-F'RTIJ: ) 0
Chapter 24 Adsorption of Inorganic Species from Aqueous Solutions
Table 24.4 Surface complexation constants of metal ion adsorption onto two activated carbon cloths, calculated using the DLM model [33]. na: not available
cu2+ Pb 2 + Ni2 +
20 40 20 40 20 40
2.909 2.102 4.137 2.545 0.068 -0.021
na
0.293 na na
-0.382 -0.606
These values are given in Table 24.4 for two initial concentrations of metal ions, 20 and 40 mg/L [33]. Values obtained for complexation constants are of the same order of magnitude as that obtained for copper adsorption onto a GAC, K:~OM+ ~ 6 [34]. 24.2.2 Adsorption of Metal Ions onto Activated Carbon
Preloaded with Organic Matter The influence of activated carbon preloaded with organic matter (see Chapter 25 for the adsorption of organics in aqueous solution) on the adsorption of metal ions is dependent on the operating conditions and on the properties of both the activated carbon and the organic matter [18]. The comparison of copper adsorption onto different activated carbons (in the form of cloth, ACC; and granules, GAC) preloaded with benzoic acid [17] shows that, whereas adsorption of Cu(II) onto loaded ACCs is greater than onto virgin ACCs, the loading of GAC by benzoic acid induces a decrease in adsorption capacity of 66 % (Fig. 24.2). As porosity and surface functional groups are similar for ACCs and GAC, this behavior difference may originate from pH conditions, initial pHo = 5 for ACC and 3.5 for GAC. At pHo = 5, adsorbed benzoic acid is in benzoate form C 6 H sCOO- (pKa = 4.2) and may form some ligands with metal cations Cu 2+, cationic species being dominant with hydrolyzed species at pH = 5. A previous work carried out with virgin activated carbon demonstrated an adsorption mechanism by ion exchange [8]. In the case of an activated carbon surface coated by benzoate ions, the following reaction may occur (where "S" is the activated carbon and "M" the metal ion), which induces an increase in metal ion removal from solution (Ce = 0 mmol/L in all cases): S-C 6 H sCOO- +M 2+ -+ S-(C 6 H sCOO-M)+.
24.2
Metal Ion Removal
+ Cu/ACC1 0.3 0.25
& Cu/ACC2 6. Cu/ACC2*BA
• Cu/GAC o Cu/GAC*BA
¢ ¢
0.2
Cu/ACC1*BA
¢
Qe (mmol/g) ¢
¢
+
¢
0.15
¢
+ ¢ ¢ +.. +. ~ ~ • •
0.1
.
+
6.
6.
~
• 0
0.05
o 0
0
•
+
•
&6..
0
• 0
0
0
0
0
0 0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Ce(mmol/L)
Figure 24.2 Adsorption of Cu(II) metal ions onto activated carbons in the form of cloths (ACC) and granules (GAC) , and onto the same activated carbons preloaded with benzoic acid [17].
The improvement in adsorption was more marked as the equilibrium concentration increased, due to the constant concentration of benzoic acid. At pHo = 3.5, adsorbed benzoic acid is in C 6 H sCOOH form and there is no ligand formation, so no increase in adsorption. The size of the organic molecules coating the activated carbon may also have an influence on the metal ion adsorption. Gold cyanide adsorption onto granular activated carbon loaded with different organic compounds (phenol, sodium ethyl xanthate, and ethanol) showed that the long-chain organic compounds had a higher degree of inhibition of gold cyanide mass transfer compared to the low-molecular-weight compounds [35]. For this reason, high-molecular-weight organic molecules, like fulvic acids, induce a reduction of Cu(II) binding by activated carbon due to either the blockage of pores or the interaction of the surface sites with fulvic acid molecules [36].
24.2.3 Saturation of Activated Carbon by Organic Matter and Metal Hydroxides After its use in wastewater treatment (see Chapter 26), activated carbon is saturated by organic matter and may contain some metal hydroxides that have been formed because of the presence of metal ions (like iron, aluminum, and calcium) in chemicals used in the coagulation and flocculation steps. Table 24.5 shows that high levels of calcium are accumulated by a field-spent activated carbon while accumulation of other metal ions occurs at much lower concentrations [37]. Usually, thermal processes using CO 2 or steam are employed for the regeneration of this field-spent activated carbon at temperatures that commonly exceed 700°C. However, the quality ofregenerated activated carbon for organics removal may be negatively affected by the presence of these metal hydroxides,
Chapter 24 Adsorption of Inorganic Species from Aqueous Solutions
Table 24.5 Ash content and inorganic composition of virgin and field-spent activated carbon Chemviron F400 [37]
M~tal 'otttent
.
~ JI r_ 1."#
Virgin GAC Spent GAC
4.8 12.6
0.089 4.045
0.618 0.480
0.467 0.543
0.026 0.029
0.069 0.061
0.020 0.062
0.002 0.006
which could catalyze the oxidation reaction between the regeneration agent (C0 2 or steam) and the solid carbon. The catalytic metals, by promoting localized oxidation of the carbon base, cause excessive mass loss due to the destruction of narrow micropores and thus a conversion of micropores into mesopores [38]. The catalytic effect of accumulated metals is especially marked with calcium [39, 40]. The presence of 2.3 % calcium increases the mass loss for steam regeneration at 850°C by a factor of 5 and for CO 2 by a factor of 25 at the same temperature compared to the respective rates observed in the absence of calcium [41]. The catalytic effect of iron was also shown on the thermal regeneration of granular activated carbon, with an increase of mass loss from 4.2 to 5.7 % in the presence of chelated Fe(III) but without an alteration of the pore structure. However, this catalytic effect of iron is no longer significant in the presence of sulfur at concentrations (0.01-0.03 %) naturally occurring in granular activated carbon [42]. To overcome calcium catalysis, a methodology has been assessed that uses specific reactivation parameters in terms of temperature ramp and duration [39, 40]. In any case, an acid washing seems to be a good way to remove the majority of metal ions before thermal regeneration of activated carbon [43].
24-3 ANION AND (ATION REMOVAL Most common ions (sodium, calcium, nitrate, phosphate, chlorine, bromide, and iodine) found in natural or wastewater are not really adsorbed onto activated carbons. An exception is fluoride that can be removed by activated carbon as well as by activated alumina. Table 24.6 gives some indications of the adsorption potential of some inorganic cations and anions onto carbonaceous porous material. Recently, the reduction by activated carbon filters of bromate (Br0 3 -), an ozone disinfection molecule produced by reactions between ozone and bromide initially present in water, has received increasing attention. Bromate reduction (Br0 3 - /Br-) was effective in a virgin activated carbon grain filter or in a BAC filter. The presence of natural organic matter (NOM) or inorganic ions (nitrate,
24.4 Reaction Between Activated Carbon and Oxidants
Table 24.6 Weak adsorption of common ions present in water onto activated carbon [2]
Adsorption potential
Low
Low
Low
Low
Low
Low
Low
High
phosphate, chloride, sulfate) caused a decrease in the mass of bromate removed by activated carbon grains as a result of the competition for adsorption-reduction sites. Increasing the empty bed contact time (EBCT), about 20 min, improves Br0 3 - removal [44-46]. In the specific case of perchlorate (CI0 4 -), a strong oxidant present in water at several micrograms per liter due to the utilization ofNH 4 CI0 4 in solid rocket fuel or components of munitions, Brown et al. [47] showed that: • in an abiotic GAC filter, percWorate was removed by ion exchange rather than by chemical reduction. The removal capacity in a column was found to be 0.172 mg/CI0 4 /g GAC. However, perchlorate was often displaced from the GAC by other ions present in the raw water. • in a biotic GAC filter, a low concentration of perchlorate was reduced biologically (less than 50 g/CIO 4 - / J..LL). The microorganisms convert perchlorate to chloride. The same operating conditions as for denitrification are required. However, this reduction was highly sensitive to nitrate. As N0 3 - concentration increased, perchlorate removal decreased. BAC filtration is a feasible option for perchlorate and nitrate treatment in drinking water. In order to improve perchlorate elimination from groundwater, granular and fibrous activated carbons preloaded with iron were used and compared to virgin GAC [48]. The activated carbon treatment was performed with ferric chloride and oxalic acid according to the procedures described in Section 24.4.1. The perchlorate adsorption capacity was found to be 0.34 mg/g in the preloaded carbon compared to 0.24 mg/g for the virgin activated carbon. The experiments were developed in a lab pilot unit and in a full-scale bed with an EBCT of 40 min. Sodium borohydride solution (100 mg/L) at about 5 % of the treated groundwater volume was able to restore the adsorption capacity of the GAC filter. The perchlorate mass flow ratio ranged from 100 to 215.
24-4
REACTION BETWEEN ACTIVATED (ARBON AND
OXIDANTS
The reactions between an oxidant and the activated carbon surface have been investigated in different ways. The first is to obtain a high level of surface functional groups to improve the interactions between these chemical functions
Chapter 24 Adsorption of Inorganic Species from Aqueous Solutions
and the pollutants (see Section 24.2 for heavy metal ion removal). The second way is to remove the residual oxidant from the aqueous solutions.
24.4.1 Direct Reaction with High Concentration Oxidants In aqueous solution, activated carbon (C), considered as a reductant, in contact with an oxidant (OX) reacts to give volatile molecules such as CO, CO 2 , H 2 0, and/or an oxidized carbon surface (C*O). This exothermal redox reaction is represented by the following general equation:
C*+OX+0 2 ~ CO+C0 2 +H 2 0+C*O Yang [49] describes some acidic surface functional groups bonded to aromatic rings present in the carbon surface. A simplified approach shows the presence of chemical groups such as carboxyl (G1), lactone (G2), phenol (G3), and carbonyl (G4). Infrared spectroscopy, X-ray photoelectron spectroscopy or ESCA, temperature programed desorption or TPD, and electrokinetic measurements have been carried out to identify and quantify these surface functional groups. However, acid-basic titration has been the most useful technique to determine the acidity (or basicity) of porous carbon [50]. More details on the surface chemistry characterization of carbons can be found in Chapter 13. Different activated carbons in contact with oxidants have been analyzed. Table 24.7 presents the oxidized material characteristics. The drastic treatments induce an increase in the chemical functions and then a better removal of heavy metal ions in aqueous solutions (Section 24.2.1). Some organic molecules, such as humic substances or more generally NOM, present in surface water have also been reported as being removed by oxidized activated carbon [51].
24.4.2 Reaction with Free Chlorine or Chlorine Dioxide One of the first mechanisms proposed for the interaction between chlorine and activated carbon is a hydrolysis, either in solution or in the adsorbed state, with the following simple reaction:
At high concentrations, free chlorine reacts directly with activated carbon to produce high-molecular-weight color-forming organics. Snoeyink et al. [52] also found chlorinated by-products in solutions such as chloroform, trichloroethane, and chlorinated aromatics. Montgomery [2] mentioned that, during a drinking water treatment, the concentration of residual chlorine in water is low and thus the level of the chlorinated compounds produced by these reactions is insignificant. Free chlorine reacts also with organic matter adsorbed onto activated carbon grains, which catalyze the reactions. Chlorinated compound precursors are, for example, NOM such as humic substances [53], proteins,
tv
~
.h-. ::::c (1) cu
Table 24.7
Characteristics of activated carbon after an oxidation reaction, Gi are the surface functional groups determined by Boehm's method, Gl for carboxylic acid, G2 for lactone, G3 for phenol, and G4 for carbonyl. These groups are differentiated by neutralization with solutions (o.oSN) of NaHC0 3 , Na 2 C0 3 , NaOH, and CH 3 CH 2 0Na [50]. 100 g of activated carbon in 500 mL of water with or without an oxidant
~
o' ::3 OJ
(1)
~ (1) (1)
::3
»
~
S
30
uQ)
.0
0en
u ca 'E ::::J 0
20
E
"0 ..........
co c>
.... E So
100
E
........ 0>
S U
Q)
.0
0en
200
u
co 'E
::J
0
E
c
'c
'(ij
0.9
MIS •
0.8
HSDMfit Geosmin • Co =39 ng/L -------- HSDMfit
e\
0.6
E ~
c
-\
0.5
\~
0
U ctS
u:
•
0.7
Co =49 ng/L
\\~,
0.4
"'~,
0.3
.
0.2
-.
0.1 0.0 0
50
100
150
200
250
Time (min)
Figure 26.8 Fraction of MIB and geosmin remaining as a function of time.
The practical application of PAC for the removal of MIB and geosmin has been aided by the application of the HSDM [63, 64, 68]. Figure 26.8 illustrates the difference in the adsorption of the two compounds, and the fit that the HSDM can give to the data. Diffusion coefficients derived from these fits can then be used to predict the adsorption of the taste and odor compounds, and consequently the PAC doses required under particular water treatment plant situations [63, 68, 69].
Chapter 26 Adsorption From Aqueous Solutions: Water Purification ..c
0)
100
::J
Cinf =50nglL MIB
e
80
..c ~
co
~
0
60
.0
c:c
40
"E Q)
20
~
0
~
Q)
0
a..
0
2
4
6
Service time (years)
(a)
..
1.99 min EBCT 2.71 min EBCT
Os =7.86 x 10-12 em 2/s
4.28 min EBCT HSDM Fit - 2.71 min EBCT
~ = 2.23 x 10-4 em/s
HSDM Prediction - 1.99 min EBCT
mc =0.4195g
HSDM Prediction - 4.28 min EBCT
Bed depth = 1.00 em
..c
0)
::J
100
~
80
~
60
e ..c co
.0
c:c ~ "E Q) ~
Q)
a..
40 20 0
0
1000
2000
3000
Time (min)
(b) Figure 26.9 Removal of MIB using laboratory GAC techniques. (a) Comparison of minicolumn results with pilot plant [17]. (Reprinted from Journal AWWA, Vol. 91, No.8 (August 1999), by permission. Copyright © 1999, American Water Works Association). (b) HSDM fits and predictions compared with SBA data. (Reproduced with permission from Re£ [68].)
Figures 26.9(a) and (b) show how laboratory techniques can assist in the application of GAC for the removal of geosmin and MIB. Figure 26.9(a) shows the results of a mini-column test on GAC removed at various intervals from a GAC pilot plant [17]. The percent removal of MIB was measured in the laboratory by running a mini-column test at the same empty bed contact time as that used in the pilot. Further details can be found in [68]. The results indicate that GAC filters could be tested at regular intervals for the removal of MIB using a simple laboratory trial. Figure 26.9(b) compares the experimental data points obtained during a SBA test with the HSDM fit and predictions [68]. The carbon had been preloaded for 2 years in a pilot plant. The predictions and the
26.3 Removal of Microcontaminants of Concern to the Drinking Water Industry
fit agree well with the data, therefore it could be expected that the results, when applied to full scale, could give a good indication of the potential performance of GAC filter for the removal of MIB.
26.3.4.2 Algal toxins Although a range of algal toxins has been identified, the most common worldwide, and therefore the most widely studied, is the microcystin group of compounds. They are cyclic compounds, consisting of seven amino acid groups, and are of around 1000 g/mol molecular weight. More than 60 variants of the microcystin toxin have been identified, differing from one another mainly in variations of two of the amino acid groups, although minor variations to the other amino acids are also seen in some variants [62]. All microcystins contain the Adda side chain, the structural unit largely responsible for the toxicity of the compounds through protein phosphatase inhibition. The most common of the over 60 known variants of the toxin, microcystin LR (mLR) , incorporates leucine and arginine in the variable positions. Although mLR is the most commonly reported of the microcystins, it is very seldom the only microcystin found in a bloom situation, and is often not present at all, with the other variants predominating. The information available in the literature on the adsorption of mLR onto activated carbon indicates that, as with the adsorption of most microcontaminants, the removal efficiency is dependent on the type of activated carbon and the water quality conditions [70-72]. Newcombe and Nicholson [73] have reported a direct linear relationship between the adsorption of the toxin and the volume ofpores between 2 and 50 nm, with a linear regression giving parameters R 2 = 0.97, P < 0.0001, and N = 9. Microcystin LR is seldom the only microcystin present in a toxic algal bloom, and in many regions mLR is not the most commonly occurring variant [74]. Very little information is available in the literature on the effect of water treatment processes on other variants. The only published investigation of the adsorption of microcystin variants other than mLR used relatively impure toxin extracts [75]. The authors suggested differences seen in the adsorption of the microcystin variants could have been due to different contaminant levels in the spiking material. The UK Water Industry Research (UKWIR) undertook a computer modeling study to compare the octanol-water partition coefficients of nine microcystin variants [76]. With this information, and molecular size data, the authors concluded that the variants should respond similarly to water treatment processes; and, in particular, that the variants would adsorb onto activated carbon to the same, or greater, extent as the commonly studied variant microcystin LR. Studies since, on four microcystin variants, have shown large differences in the adsorption of the variants [71, 77, 78]. Surprisingly, the largest and the most hydrophilic of the toxins studied, mRR with two arginine groups in the variable positions, showed the highest adsorption on a range of PACs, contrary
Chapter 26 Adsorption From Aqueous Solutions: Water Purification
700
to expectations for adsorption onto activated carbon. The reason for this effect is still not clear. The adsorption of microcystins has been shown to be strongly affected by NOM [78]. In this work the effect, for four microcystin variants, was shown to be a function of the DOC concentration. This is probably a result of the direct competitive effect, where the competitive NOM, those compounds approximately the same molecular weight as the microcystins, makes up the bulk of the NOM. Therefore the bulk characterization parameter, DOC, gives an indication of the concentration of competing compounds, where for MIB and geosmin it could not [27, 63, 69]. Figure 26.10 displays the large difference in the adsorption ofmLR and microcystin LA (mLA) as a function of time. Microcystin LA (comprising alanine, rather than arginine as one of the variable groups) is a smaller, more hydrophobic molecule than mLR, but adsorbs to a much lower extent. The HSDM was used to fit the data, and predict kinetics of adsorption under different conditions. As can be seen from the figure, the predictions are excellent. Computer modeling of kinetics of adsorption was used to predict the PAC doses required to reduce the two compounds to below the WHO guideline of 1.0 f.1g/L in 60 min. The results are given in Table 26.3. Clearly, the differences in adsorption between microcystin variants will have a significant effect on treatment options available to water suppliers. Some previous studies showed that GAC is effective in the removal of the microcystin hepatotoxins [79, 80]. These studies were undertaken on new, or virgin, activated carbon, which has a very large pore volume available for
mLA
mLR 1.1
•
1.0
-
C)
c
'c '(ij
E ~
\
0.9
~
\
1.0 0.9
-. -.~.-.-.
HSDM prediction
\
0.8
Co =17.3 mg/L Cc =15 mg/L HSDMfit Co =14.0 mg/L Cc =25 mg/L
\
....
0.8
•. •
ti~
0.7
e
0.6
0.7
() ()
'E c
0.5
~
0.4
"-
T '. ,
•
T.,
0
~
....
....
\
C
-
• ...
Co =19.2 mg/L Cc =15 mg/L
HSDMfit Co =15.6 mg/L Cc =25 mg/L _. _.. HSDM prediction
0.6 0.5
',T.-
T T
u.
0.4
0.3 0.3 0.2 0
10
20
30
40
Co = initial concentration
50
60
0
10
20
30
40
50
60
Time (min)
Cc=PAC dose
Figure 26.10 Adsorption ofmicrocystins LR and LA as a function of time, showing HSDM fits and predictions. (Reproduced with permission from Ref. [71].)
26.3 Removal of Microcontaminants of Concern to the Drinking Water Industry
701
Table 26.3 Predicated PAC doses required to obtain a concentration of 1 J-Lg/L after 60 min contact
38
10 5 2
»100 95 50
29 15
adsorption and is capable of adsorbing a wide range of organic compounds. When similar studies were conducted over a longer timescale, or on activated carbon which had been in use for a period, it was found that breakthrough of toxin can occur in the eilluent from the filter [81, 82]. This is most likely due to both direct competition for adsorption sites and fouling of the GAC surface by NOM compounds of approximately the same size as the microcystins. As mentioned above, this NOM represents the majority of the DOC, and thus results in significant fouling effects. It has also been shown that breakthrough occurs at a different rate, depending on the DOC concentration of the inlet water [82]. Predictive tools have been used in an attempt to predict breakthrough of microcystins, but in these cases the results were not accurate due to biological degradation occurring, as well as adsorption [83, 84]. Figure 26.11 illustrates both forms of removal on a laboratory-scale GAC column. Initially the GAC,
.-.-.-.-.-==---.-.-..- ..-.-.
Sterilization
100
80
,.ll~'J/~~
~
o>
E ~
II1\ 60
/ \ /;
..
c
\,'
·x .9 C Q) ~
."w
.1I
uQ)
I
•/\
I
II
/ \
/
\ I\ ~
40
\: II"
.~
/ \/ \ I \I
\
Q)
a..
•
20
I
\/ •
•
1-_-mLR I -.-mLA
l II
•
Adsorption
+llII-----------------1.~~---~
Biodegradation
Adsorption
O-+----,--.....__----r--.-----r---,-~----r-.....__----r--~-,.---,r----+-____r-....,..........---.
o
10
20
30
40
50
60
70
80
Days
Figure 26.11 Adsorption and biodegradation taking place on a GAC laboratory column.
Chapter 26 Adsorption From Aqueous Solutions: Water Purification
702
preloaded with NOM, adsorbed both toxins to a certain extent. The fact that mLR was removed to a greater extent than mLA indicated the removal mechanism was mainly physical adsorption. However, after an "acclimation" period of 16 days, the toxin was no longer detected in the effiuent from the column. When the GAC was removed from the column, sterilized and replaced, adsorption was again the main mechanism. This is an excellent illustration how physical and biological processes working together can optimize the use of GAC.
26.4
REMOVAL OF NATURAL ORGANIC MATERIAL
As mentioned earlier, activated carbon is not commonly employed specifically to remove NOM. When PAC is applied, the removal of NOM is usually very low compared with the bulk DOC concentration. However, the removal of NOM through GAC filters can be high when the filter is first commissioned, and can reach a steady state removal ofbetween 10% and 20%. This is often considered to be due solely to biodegradation of the more readily degradable NOM, but is probably due to a combination of physical adsorption and biological processes. The removal of NOM, even at relatively low levels, is advantageous to the water supplier as it results in lower formation of disinfection by-products, as well as lower consumption of disinfectant, and a longer maintenance of the disinfectant residual in the distribution system. Due to this fact, and the significance of NOM as a competitor in the removal of other target compounds, a great deal of literature has been published on the adsorption behavior of N OM onto activated carbon. Many workers have shown that the pore size distribution has a significant effect on the adsorption of NOM [5, 6, 53, 85, 86] with some surface chemistry effects also noted [7, 87, 88]. Unlike the contaminants mentioned previously in this chapter, NOM can be highly charged, therefore the surface charge and solution conditions can have a significant effect on the adsorption of NOM. Several authors have shown that electrostatic effects, both between adsorbed NOM molecules, and between the carbon surface and the adsorbate, influence the adsorption, perhaps to as great an extent as the pore volume [6,7,31]. Figure 26.12 shows combined data for 10 activated carbons (described elsewhere [89, 90]), and two different NOM molecular weight fractions [91]. The average hydrodynamic diameters for each NOM fraction were measured using flow field flow fractionation and these values were related to the pore volume available for adsorption in that size range. The amount of DOC adsorbed at a solution concentration of DOC = 100 mg/L was plotted against available pore volume. The results show that the adsorption of NOM, at pH 3, is governed solely by the size of the adsorbate, and is due to dispersion forces, unaffected by the surface chemistry of the carbon. On the other hand, the adsorption of NOM at pH 7, also shown in Fig. 26.12, cannot be attributed to the available pore volume alone. Bj elopavlic et al. [92] showed that electrostatic effects between the surface and the adsorbate were important at low surface
703
26.5 Conclusions
400
'Oi .........
C)
•
pH 3, R=0.99, P.
()
.........
c
Q)
~
:::J
c
C'"
:J
U.
~
Q)
°C\loq-COCOOC\loq-cocoo ..... C\I (Y) oq- co f'o. co C'l 0 C\l 0 0 0 0 0 0 0 0 ..........
CT
~
0000000000
LL
Cross-sectional
arealmm 2
oq-oq-oq-oq-oq-oq-oq-oq-oq-oq-oq-oq-oq-oqC\i~ ~
0)
.::£
Q)
80
.::£
.........
~
~
~
'0 1: ::J
"0 ~
60
~ ()
0
E-
(j)
Fir fibers carbonized at 380°C
•
Exfoliated graphite
•
50
~
40 30
0 (j) 20
DO
20
0
a0
Fir fibers carbonized at 900°C
•
[]
c
40
E c
60
a.
O
~
[J
o
0)
"'C 0)
70
.........
Q)
..c.
[J
10 0
~'b ..
[J
0
0 0
25
50
75
100
125
0
20
3 Bulk density (kg/m )
40
60
80
100
Bulk density (kg/m 3)
(c) Carbon fiber felts
90
~
80 ~
0)
.........
~
"0 ~
a. ~ ()
c
o
Carbon fiber felts
70
•
Firfibers
60
•
Exfoliated graphite EG-1
-,
50 40
0
a0
(j)
~
30 20 10 0 0
~~-
0-
20 40 60 80 100 120 140 160 180 200 Bulk density (kg/m 3)
Figure 27.6 Dependences of sorption capacity for the A-grade heavy oil on bulk density of carbon sorbents.
In Fig. 27.8(a) and (b), the bulk density dependences of the sorption capacity for the viscous C-grade heavy oil, whose viscosity is 0.35 Pa s, are shown on exfoliated graphite and carbonized fibers. In Fig. 27.8(a), the sorption capacity of the B-grade heavy oil with a little lower viscosity (0.27 Pa s) is also plotted. Sorption capacity for the viscous C-grade oil is relatively low and its decrease \vith increasing bulk density is observed markedly for the C-grade than for the less viscous A-grade. No sorption was detected on dense exfoliated graphite, but on carbonized fir fibers a certain capacity could be observed. Sorption capacity
27.3 Sorption Capacity for Viscous Organics
o
100
50
o
Exfoliated graphite
~ Carbonized fir fibers
•
Activated carbon
[] Carbon fiber felt
O - - _......_ _......._ _......II...--._ _
o
50
.....-_-~
100
Pore volume (ml/g)
Figure 27.7 Relation between sorption capacity ofdifferent carbon sorbents for the A-grade heavy oil and pore volume measured by using a new dilatometer for mercury porosimetry.
(a) Exfoliated graphite
(b) Carbonized fir fibers 70 ~
100 [] B-grade heavy oil
0) ~
..........
0'>
\
75
~ ~
a. (lj ()
40
a 0
CJ)
C A-grade heavy oil
20
•
Crude oil
1\ B-grade heavy oil
•
0 0.001
~ ~
'0
C-grade heavy oil
0.01
0.1
~ 990 '0.",
III
. . o"~11
60
o
.~
It
(lj
\
a. (lj ()
~
0
•
80
0,
~~
•
60
0> .::::t:.
'~
(lj
c:
,•
100
\ 1
c:
40
It
0
a 0
CJ)
10
\
20
0 0.0001
Viscosity 'TJ (Pa s)
0.001
0.01
0.1
10
Viscosity 'TJ (Pa s)
Figure 27.9 Dependences of sorption capacity of exfoliated graphite with a bulk density of 7 kg/m3 on the viscosity of various oils.
of heavy oils, based on the measurements of sorption capacity at different temperatures from O°C to 30°C. Figure 27.9(a) shows a strong dependence of sorption capacity on viscosity of sorbate heavy oils, and their dependence seems to be divided into two; low viscosity region for the A-grade and crude oils and high viscosity region for the B- and C-grade oils. However, the measurements on a wide range of viscosity using various oils show that sorption capacity of the exfoliated graphite levelled off in the intermediate viscosity range, as shown in Fig. 27.9(b), and will be discussed in the next section.
27.3.2 Various Oils Other than Heavy Oils Various oils were selected to cover a wide range ofviscosity as summarized in Table 27.2. The exfoliated graphite used was the same one as above, whose bulk density was about 7 kg/m3 •
Table 27.2
Oils used in the present work
Kerosene Light oil A-grade heavy oil Crude oil Mineral oil Grape seed oil Salad oil
788.3 823.2 852.3 867.7 842.5 919.9 918.3
0.001 0.001 0.004 0.007 0.033 0.052 0.056
Saillower oil Two-cycle motor oil Four-cycle motor oil Diesel oil B-grade heavy oil C-grade heavy oil
914.0 858.3 876.1 877.7 902.7 925.5
0.069 0.118 0.126 0.127 0.160 0.350
27.3 Sorption Capacity for Viscous Organics
721
In Fig. 27.9(b), the sorption capacity is plotted against viscosity 11. The sorption capacity that was determined by direct soaking of an exfoliated graphite lump with a bulk density of 7 kg/m3 into oils at different temperatures show a marked increase at the high-viscosity side and a marked decrease at the lowviscosity side. The sorption capacity value, which levels off at intermediate viscosity, is around 70 gig.
27.3.3 Biomedical Fluids Biomedical molecules are typically large molecular and weak polar materials, for which exfoliated graphite has a large sorption capacity. The sorption performance of exfoliated graphite with different bulk densities was studied on several kinds ofbiomedical molecules, ovalbumin, serum albumin, bovine serum albumin (BSA), lysine, and herring sperm DNA [19]. Changes in absorbed amount of BSA with time showed that sorbed amount increases quickly in the first 20 min, but exfoliated graphite with the higher bulk density quickly reaches saturation at the lower amount. The lower density sample has larger sorption capacity, because ofits larger pore volume for capillary condensation and takes longer time to be saturated. In Fig. 27.10, dependences of sorption capacity on bulk density of exfoliated graphite are compared with three liquids - heavy oil, gasoline, and BSA. Sorption capacity decreases with increasing bulk density of exfoliated graphite for the three liquids, because of the decrease of pore volume, particularly the decrease of large spaces among the particles, in the lump of exfoliated graphite. The characteristics of carbon materials, such as low weight, chemical inertness, excellent compatibility with the human body [29], and also bacteriostasis, led to the new application namely: medical dressing in preventing a traumatized
70 _
..,....----------------t
60
0) -.... -9 50 C
\
~ 40
\
«j ()
c:
o
30
aa 20
\,
\Heavyoil
0
~Gasolin~
-~:q
CJ)
O'Q......
~~[j--D
"'0
1), Yo the average pore radius of the porous sorbent, and () the contact angle of interface between the liquid and the pore wall of sorbent. Thus, in the first bracket in Eqn (27.3) are the parameters based on the sorbates, the second one are the variables for sorbents, and the third one is the parameter for the interface between the sorbate and the sorbent. In order to understand the dependences of the sorptivity Ks on the bulk density of carbon sorbents for the A-grade heavy oil, the first bracket in Eqn (27.3) must be a constant, and also the third bracket can be assumed to be a constant because the contact angle () between the A-grade heavy oil and the pore wall of carbon is not supposed to have much difference among the carbon materials used. Therefore, the sorptivity Ks can be approximated to depend on the three parameters of the carbon sorbent, effective sorption porosity e*, average tortuosity factor A, and average pore radius Yo. The effective sorption porosity, e*, is the volume ratio of macropores directly involved in the sorption and e* is calculated using the value m7 [32], at which the linear relation between ms vs (1/2 is broken away (refer Fig. 27.13) from the equation:
e * =m7 -, Ld}
(27.4)
where L is the height of sorbent. With increasing bulk density of sorbent, the effective sorption porosity e* increases rapidly, but the average pore radius Yo decreases. The tortuosity A is mainly governed by the smoothness of the surface of sorbent particles. In order to have a high sorptivity Ks ' therefore, a low value of A, i.e., smooth surface of particles, and high bulk density where changes in both e* and Yo becomes small, is desired. This occurs in carbon fiber felts among the three carbon sorbents used, as shown in Fig. 27.14. In the case of exfoliated graphite, e* shows a maximum at a bulk density of about 16 kg/m3 , pore size distribution changes markedly, i.e., the average pore radius Yo decreases with increasing bulk density, and the value of A is expected to be the largest among the three sorbents.
Chapter 27 Sorption of Viscous Organics by Macroporous Carbons
72 6
Therefore, the complicated change of K s observed on exfoliated graphite seems to be a result of balancing between the dependences of 8* and Yo on the bulk density of the exfoliated graphite [21]. In the case of carbonized fir fibers, however, a marked dependence of K s on their bulk density was observed, from a low value comparable to exfoliated graphite to a high value comparable to the carbon fiber felts. Their Ks seemed to be mainly governed by 8*, which has also pronounced dependence on bulk density, even though the average pore radius Yo seems to become small with increasing bulk density. For the carbonized fir fibers with high bulk density, 8* tends to be saturated and pore size distribution becomes simple, which seems to lead the value of K s to be a constant near that for carbon fiber felts [11, 21]. A detailed discussion on the dependence of K s on the bulk density of carbon sorbents was presented in our papers [20, 21]. In Fig. 27.15, sorption curves observed on the three heavy oils - the A-, B-, and C-grade oils - and salad oil are compared in order to show the effect of viscosity of oils on their sorption rate. The sorption rate depends strongly on viscosity. Less viscous oil (e.g., A-grade heavy oil and salad oil) reaches saturation very quickly. However, viscous oil (e.g., C-grade heavy oil) is sorbed very slowly into exfoliated graphite. On the other hand, saturated amounts of sorbed oil for the three oils - the A- and B-grade heavy oils and salad oil - reached almost the same value after about 1 h (about 3600 s). For the viscous C-grade heavy oil, however, it took a long time, about 24 h, to reach a saturation, but the saturated amount of oil was a little less than those for other less viscous oils.
5.0
r------~--------------,
A-grade heavy oil
Salad oil
4.0
§ en en ctS E
3.0
"C
Q)
o
.0
2.0
en
00
1.0 0°
0 0
0
00 0 C-grade heavy oil
OL-.-l.--.-I---'---'--..L---l._--L---'-~--'-~~"'--I~"-""-~
o
500
1000 Time (s)
Figure 27.15 Sorption curves for different oils.
1500
2000
727
27.5 Recovery of Heavy Oils
8 •
Kerosene
7
-
~
6
en
..........
C\I
E
5
.......... C)
~ ~rn
~ 'S;
a 0
(j)
4 3
2
Salad oil ~ Safflower oil Grape seed '~-~YCle ~otor oil 2-cycle motor oil Diesel 011 . C-grade heavy 011 OI...o-....................--"""..A.Ao......u - .......................---~.....--
0.0001
0.001
0.01
0.1
Viscosity J-L(Pa s)
Figure 27.16 Dependence of sorptivity Ks on the viscosity of oils.
In Fig. 27.16, sorptivity K s is plotted against viscosity JL of oils in logarithmic scale. Ks shows a strong dependence on JL; the oil with the higher viscosity is sorbed into a column of exfoliated graphite with the slower rate.
27.5
RECOVERY OF HEAVY OILS
For heavy oils, their spillage by accidents result in not only the contamination of the environment but also great loss of energy resources. Therefore, their recovery from sorbents is also an important problem to be solved. Recovered heavy oils have to be usable as energy resources and also recycling of the sorbent carbons is strongly desired. From this point of view, cyclic performance of carbon sorbents was examined by different processes: filtration under suction, washing by solvent for heavy oils, centrifugation, etc. Less viscous oils, such as the A-grade and crude oils, could be recovered from all carbon sorbents by a simple filtration under suction. The changes in amounts of sorbed and recovered oils with cycling are shown on the A-grade oil in Fig. 27.17(a-c). On exfoliated graphite, about half of the sorbed oil can be recovered in each cycle of sorption and recovery. The remaining oils in the exfoliated graphite lump disturb the further sorption and so the sorption capacity of the lump decreases roughly by half for each cycle. This decrease in sorption capacity is reasonably supposed to be due to the oils that are trapped in the small crevice-like pores and also in the pores inside the particles.
728
Chapter 27 Sorption of Viscous Organics by Macroporous Carbons
(a) Exfoliated graphite with 7 kg/m 3
(b) Carbonized fir fibers with 5.5 kg/m
3
50
en ::t:. .......... C)
C. .!!2 '0 u (])
~
40
-
c.> ~
Sorption
0
Recovery
~
30
.....
0
II
~
~
20
~
u
(])
.0
10
0 en 0 1st
2nd
3rd
4th
5th
1st 2nd 3rd 4th 5th 6th 7th 8th
Cycling time
Cycling time
(c) Carbon fiber felt with 72.6 kg/m 12 . - - - - - -
3
,.......--.---,..---~--.
1st 2nd 3rd 4th 5th 6th 7th 8th
Cycling time
Figure 27.17 Cyclic performance of carbon sorbents for the A-grade heavy oil by filtration under suction.
The performance of sorption/recovery of carbonized fir fibers by filtration under suction is much better than exfoliated graphite as shown in Fig. 27.17 (b). In each cycle, about 80% of sorbed oil is recovered and so the decrease in sorption capacity with cycling is much slow, after eight cycles it becomes about 60% of that of the first cycle. When the fiber lump with a high bulk density was used, the decrease in sorption capacity with cycling is much less, though the absolute value of sorption capacity is less. Although sorption capacity could not be high, the cycling performance of carbon fiber felts was excellent. By filtration under suction, about 90% of the A-grade heavy oil sorbed could be recovered and no reduction in the sorption capacity was observed even after eight cycles, as shown in Fig. 27.17(c).
27.5 Recovery of Heavy Oils
72 9
Viscous oils could not be recovered from either exfoliated graphite or carbonized fir fibers by filtration even under a strong suction. Sorbed heavy oils could be recovered by washing with a solvent, such as n-hexane, but the exfoliated graphite after washing could not be reused as sorbent for heavy oil, mainly because of the destruction of the bulky texture of the exfoliated graphite. From fir fibers and carbon fiber felts, however, the oils, even the viscous C-grade oil, could be washed out using a solvent. In Fig. 27.18(a) and (b), the cycling performances of the carbonized fir fibers and carbon fiber felts, respectively, the for A- and C-grade heavy oils by washing with n-hexane are shown. For the carbon fiber felts, almost 100% recovery and excellent cyclability by washing with n-hexane were obtained for both the A- and C-grade heavy oils. In the
60
(a) Carbonized fir fibers A-grade heavy
01
C-grade heavy oil
I-
Oi
~ 0>
50
-
~
Oi ~
-
0> ~
.J!2 ·0 40
-
'""""
u
·0 l-
CD
u
I-
CD
CD 30 > 0 ~
20
0
10
u CD .c
o
~ 20
~
u
~ 10
o
CJ)
CJ)
o
0
1st 2nd 3rd
4th
5th
6th
7th
8th
II
Cycling time
o
(b) Carbon fiber felt 14 A-grade heavy oil I_ ~
Oi ~ 0>
30
~
I-
()
~
40
.J!2
1st
2nd
3rd
4th
5th
Cycling time
Sorption Recovery
C-grade heavy 01
25
- -
12
~
0.-
..-
~
I-
..-
I-
.J!2 10
·0
-g ~
8
8
6
~
4
~
u
CD
~
2
CJ)
1st 2nd 3rd 4th
5th
6th 7th
Cycling time
8th
o 1st 2nd 3rd 4th 5th 6th 7th 8th
Cycling time
Figure 27.18 Cyclic performance of carbonized fir fibers and carbon fiber felt for the A- and C-grade heavy oils by washing with n-hexane.
Chapter 27 Sorption of Viscous Organics by Macroporous Carbons
730
case of the C-grade heavy oil, the less viscous A-grade oil can be used as a solvent, this washing process corresponds to the preparation of the B-grade heavy oil. In the case of carbon fiber felts consisting of PAN-based carbon fibers, even centrifuging with 3800 rpm could be applied without any reduction of sorption capacity during the eight cycles (Fig. 27.19). A fair amount of cyclablity was obtained even by squeezing the felt of PAN-based carbon fibers. For the recovered oils, different analyses on chemical composition, hydrocarbon contents, and molecular weights were carried out. No appreciable difference was detected between original and recovered oils. In Table 27.3, the results on the fraction of aromatic hydrocarbons and different molecular weights are summarized for the A-grade, crude, and C-grade oils. These experimental results showed that the recovered oils could be used for all applications.
(b) Bulk density of 85.5 kg/m 3
(a) Bulk density of 63.6 kg/m 3 _
II Sorption
14
C>
.::£
~
~
0, 12
I"-
-
I"-
-
DRecovery -
Oi
14
C>
~
·0 10
!!l.
!!l. ·0 10
"0
"0
~ () ~
~
8
~
6
()
~ "0
"0
~
.eo
c:
4
as
-
12
~
~
-
.::£
.........
8
6
4
"0
"0
Q)
2
..c
en 0
en
o
2
1st 2nd 3rd 4th 5th 6th 7th 8th
1st 2nd 3rd 4th 5th 6th 7th 8th
Cycling time
Cycling time
Figure 27.19
Cyclic performance of carbon fiber felts by centrifuging with 3800 rpm.
Table 27.3
Fractions of aromatic hydrocarbons and averaged molecular weight values measured by field desorption mass spectrometry (FD-MS) analysis on the crude and (-grade oils
1.06 1.06
A-grade heavy oil Original Recovered
4.0 4.2
258 259
274 274
Original Recovered
4.9 4.5
645 672
869 915
1102 1147
1.35 1.36
1.27 1.25
C-grade heavy oil Original Recovered
5.4 4.6
1071 1207
1768 1839
2428 2393
1.65 1.52
1.37 1.30
Crude oil
Farom - fraction of aromatic hydrocarbon, M n - number-averaged molecular weight, Mw - weight-averaged molecular weight, Mz - z-averaged molecular weight.
731
27.6 Discussion
27.6
DISCUSSION
So far, mats of some polymers, such as poly(propylene) and poly(urethane), have been used for the sorption of spilled oil. Their maximum sorption capacity is about 10-30 g of heavy oil per 1 g ofpolymer [35]. However, they sorb water, as well as heavy oil, and show no special selectivity for heavy oils. Therefore, the effective sorption capacity of the polymer mats for heavy oils floating on water must be lower than the figures mentioned above. Some natural sorbents prepared from cotton fibers, milkweed flosses, and kenaf plants were reported to have rather high sorption capacity and certain potential for oil recovery and sorbent reusability [35-41]. The sorption capacity of macroporous carbon materials, exfoliated graphite, and carbonized fir fibers, is very high in comparison with these materials. Preferential sorption of oils is an advantage of carbon materials in addition to their high sorption capacity. It is interesting to point out that most materials, which have either been used or tested for sorption of heavy oils, are composed from fibrous particles, as explained above. Carbon materials, which had interesting results for heavy oil sorption, are also fibrous, worm-like particles in exfoliated graphite having also fibrous morphology. The reason for this is not clear yet, but easy formation of large spaces with appropriate size for heavy oil sorption associated with easy deformation of fiber networks to give appropriate morphology to keep oils might be one of the reasons. For the large sorption capacity of carbon materials, large spaces among fibrous particles are reasonably supposed to be responsible. In the case of the lump of exfoliated graphite, there are at least three kinds of pores - large spaces among entangled worm-like particles with fibrous morphology, crevice-like pores on the surface of worm-like particles, and elliptic pores inside the particles. The large spaces among the particles occupy about 75% of the total volume of the lump of exfoliated graphite and about 70% of the sorbed heavy oil fill these spaces [23]. These large spaces among the particles can be easily destroyed by a slight compression, and, as a consequence, the sorption capacity drops down correspondingly, as shown in Fig. 27.3. However, the other two pores, the crevice-like pores on the surface of particles and the elliptic pores inside the particles, also have important roles for heavy oil sorption. Observations under the optical microscope showed that the oil rose through the edge of the crevices formed on the surface of the particles immediately on pumping heavy oil into the lump of exfoliated graphite [16]. This complicated pore structure in the exfoliated graphite lump may result in rather strong holding of sorbed heavy oils, which did not move to the filter paper during filtration to recover from the water surface, even though sorption rate is low in comparison with carbon fiber felts that have a smooth surface. The same discussion on heavy oil sorption into carbonized fir fibers is reasonably assumed, which have similar pore structure, large spaces among the fibers of fir plants, small pores inside the fibers, and also rough surface of the fibers. In
732
Chapter 27 Sorption of Viscous Organics by Macroporous Carbons
carbon fiber felts, however, only interparticle pores exist, which seems to result in small sorption capacity, but high sorption rate and high recovery ratio. Hydrophobic (oleophilic) nature of the surface of carbon materials seems also to be a factor governing heavy oil sorption, particularly for preferential sorption of heavy oils.
27.7
CONCLUSIONS
All experimental results on the sorption of viscous organics, such as various oils including heavy oils and biomedical fluids, by macroporous carbon materials [2-24] revealed that the sorption of large amount of viscous organics into carbon materials is due to the capillary pumping based on their pore structure; capillary pumping is assisted by intraparticle pores, such as crevice-like on the surface of worm-like particles and ellipsoidal pores in the particles of exfoliated graphite, and most of the organics pumped up are kept in the large interparticle spaces. Certain possibilities of these macroporous carbon materials to be used for the protection of environment from heavy oil pollutions and the reuse of spilled heavy oils were demonstrated.
ACKNOWLEDGMENTS
This series of works on heavy oil sorption and recovery was carried out in the Proposal-Based New Industry Creative Type Technology R&D Promotion Program of New Energy and Industrial Technology Development Organization (NEDO), Japan (No. 98Ec-12-002), and under the Joint Research Program betweenJapan Society for the Promotion of Science aSPS) and National Natural Science Foundation of China (NSFC). The works were partly supported by a grant of Frontier Research Project from Ministry of Education, Japan.
REFERENCES
1. Fujiraito Ind. Co., Ltd. (1979). Japanese Patent Proposal (No. 95333). 2. Cao, N.Z., Shen, W.C., Wen, S.Z., et al. (1996). The adsorption performance of heavy oil on expanded graphite. Carbon '96, New-castle upon Tyne, UK, pp. 114-15.
References
733
3. Toyoda, M., Aizawa, J., and Inagaki, M. (1998). Sorption and recovery of heavy oil by using exfoliated graphite. Desalination, 115, 199-201. 4. Toyoda, M., Moriya, K., and Inagaki, M. (1999). Sorption of heavy oil into exfoliated graphite-influence of bulk density and pore for sorption. TANSO, 187, 96-100 (in Japanese). 5. Toyoda, M., Moriya, K., Aizawa, J., et al. (2000). Sorption and recovery of heavy oils by using exfoliated graphite. Part I: maximum sorption capacity. Desalination, 128, 205-11. 6. Inagaki, M., Konno, H., Toyoda, M., et al. (2000). Sorption and recovery of heavy oils by using exfoliated graphite Part II: recovery of heavy oil and recycling of exfoliated graphite. Desalination, 128, 213-18. 7. Tryba, B., Kalenczuk, R.J., Kang, F., et al. (2000). Studies of exfoliated graphite (EG) for heavy oil sorption. Mol. Cryst. Liq. Cryst., 340, 113-19. 8. Tryba, B., Morawski, A.W., Kalenczuk, R.]., and Inagaki, M. (2003). Exfoliated graphite as a new sorbent for removal of engine oils from wastewater. Spill Sci. Technol. Bull., 8, 569-71. 9. Inagaki, M., Kawahara, A., and Konno, H. (2002). Sorption and recovery of heavy oils using carbonized fir fibers and recycle. Carbon, 40, 105-11. 10. Inagaki, M., Kawahara, A., and Hayashi, T. (2001). Sorption, recovery and recycling of heavy oil by using carbonized fir fibers. Res. Rep. Aichi Inst. Technol., 36, 69-78 (in Japanese). 11. Inagaki, M., Kawahara, A., Iwashita, N., et al. (2002). Heavy oil sorption and recovery by using carbon fiber felts. Carbon, 40, 1487-92. 12. Inagaki, M., Shibata, K., Setoh, S., et al. (2000). Sorption and recovery of heavy oils by using exfoliated graphite part III: trials for practical applications. Desalination, 128, 219-22. 13. Toyoda, M., Dogawa, N., Seki, T., et al. (2001). Sorption and recovery ofA-grade heavy oil by using exfoliated graphite packed in plastic bag - trial for practical applications. TANS 0, 166-9 (in Japanese). 14. Toyoda, M. and Inagaki, M. (2000). Heavy oil sorption using exfoliated graphite. New application of exfoliated graphite to protect heavy oil pollution. Carbon, 38, 199-210. 15. Inagaki, M., Toyoda, M., and Nishi, Y. (2001). Sorption, recovery and recycling of heavy oils by carbon materials. Kagaku Kougaku, 65, 179-82 (in Japanese). 16. Inagaki, M., Toyoda, M., Iwashita, N., et al. (2001). Exfoliated graphite for spilled heavy oil recovery. Carbon Sci., Korea, 2, 1-8. 17. Inagaki, M., T oyoda, M., Iwashita, N., et al. (2002). Sorption, recovery and recycle of spilled heavy oils using carbon materials. TANS 0, 16-25 (in Japanese). 18. Toyoda, M. and Inagaki, M. (2003). Sorption and recovery of heavy oils by using exfoliated graphite. Spill. Sci. Technol. Bull., 8, 467-74. 19. Kang, F., Zheng, Y.P., Zhao, H., et al. (2003). Sorption of heavy oils and biomedicalliquids into exfoliated graphite - researches in China. New Carbon Mater., 18, 161-73. 20. Nishi, Y., Dai, G., Iwashita, N., et al. (2002). Evaluation of sorption behavior of heavy oil into exfoliated graphite by wicking method. Mater. Sci. Res. Int., 8,43-8. 21. Nishi, Y., Iwashita, N., Sawada, Y., and Inagaki, M. (2002). Sorption kinetics of heavy oil into porous carbons. Water Res., 36, 5029-36.
734
Chapter 27 Sorption of Viscous Organics by Macroporous Carbons
22. Nishi, Y., Iwashita, N., and Inagaki, M. (2002). Evaluation of pore structure of exfoliated graphite by mercury porosimeter. TANS 0, 31-4 (inJapanese). 23. Zheng, Y.P., Wang, H.N., Kang, F.Y., et al. (2004). Sorption capacity of exfoliated graphite for oils - sorption in and among worm-like particles. Carbon, 42,2603-7. 24. Toyoda, M., Nishi, Y., Iwashita, N., and Inagaki, M. (2002). Sorption and recovery of heavy oils by using exfoliated graphite. Part IV: discussion on high oil sorption of exfoliated graphite. Desalination, 151, 139-44. 25. Inagaki, M. and Suwa, T. (2001). Pore structure analysis of exfoliated graphite using image processing of scanning electron micrographs. Carbon, 39, 915-20. 26. Kang, F.Y., Zheng, Y.P., Wang, H.N., et al. (2002). Effect of preparation conditions on the characteristics of exfoliated graphite. Carbon, 40, 1575-81. 27. Inagaki, M., Tashiro, R., Toyoda, M., et al. (2004). Pore structure of exfoliated graphite prepared from residue compounds with sulfuric acid. J. Ceram. Soc. Jpn., 112, S1513-16. 28. Inagaki, M., Tashiro, R., Washino, Y., and Toyoda, M. (2004). Exfoliation process of graphite via intercalation compounds. J. Phys. Chem. Solids, 65, 133-7. 29. Bokros, J., LaGrange, L.D., and Shoen, FJ. (1972). Control of structure of carbon for use in bioengineering. In Chemistry and Physics of Carbon, Vol. 9 (P.L. Walker, ed.). New York: Marcel Dekker, pp. 103-71. 30. Cao, N.Z., Shen, W.C., Wen, S.Z., et al. (1996). The adsorption of proteins on expanded graphite. Extended Abstracts of the European Conference, Carbon '96, Newcastle upon Tyne, UK, pp. 258-9. 31. Aggarwal, R. (1977). Evaluation of relative wettability of carbon fibers. Carbon, 15,291-3. 32. Washburn, E.W. (1921). The dynamics of capillary flow. Phys. Rev., 17,273-83. 33. Beltran, V., Escardino, A., Feliu, C., and Rodrigo, M.D. (1988). Liquid suction by porous ceramic materials. Br. Ceram. Trans. J., 87, 64-9. 34. Beltran, V., Barba, A., Rodrigo, M.D., and Escardino, A. (1989). Liquid suction by porous ceramic materials: 2. Influence of pressing conditions. Br. Ceram. Trans. J., 88,219-22. 35. Chol, H.M. and Cloud, R.M. (1992). Natural sorbents in oil spill cleanup. Environ. Sci. Technol., 26, 772-6. 36. Drelich, J., Hupka, J., and Gutkowski, B. (1988). Absorptivity of fibrous mats applied for removing spilt oil. Chemistryfor Protection of the Environment 1987, Studies in Environmental Science. Elsevier, 34, pp. 207-21. 37. Johnson, R.F., Manjrekar, T.G., and Halligan, J.E. (1973). Removal of oil from water surfaces by sorption on unstructured fibers. Environ. Sci. Technol., 7, 439-43. 38. Yamamoto, H. (1998). Manufacturing of oil sorbent from heat treated wood fiber and developing new products. Cellulose Commun., 5, 148-51. 39. Miyata, N. (1999). Oil sorbency ofsorbents prepared from kenaf (Hibiscus cannabinus L.) Plants. Sen'i Gakkaishi, 55, 576-83. 40. Umehara, K., Nakamura, S., and Saito, M. (1997). Sorbents for oils derived from woods. 27th Symposium on Chemical Treatment of Woods, Proceedings, pp. 4957 (in Japanese). 41. Inagaki, M., Nagata, T., Suwa, T.,et al. (2004). Sorption kinetics of various oils into exfoliated graphite. Fresenius Environ. Bull. (in press).
SUBJECT INDEX
activated carbon, 40 acid-basicity, 656 adsorption capacity, 634 influence of operation conditions, 635 catalytic reactions with, 645 caustic-impregnated, 535 for supercapacitors, 609 granular (GAC), prediction of adsorption behavior, 684 ignition of, 551 metal impregnated, 545, 645 nitrogen-containing, 541, 544 photocatalysis with, 646 powdered (PAC), prediction of adsorption behavior, 684 preloading with organic matter, 638 preparation by KOH activation, 610 reaction with oxidants, 641 saturation, 639 surface oxygen complexes in, 539, 544, 547, 584, 642, 645, 657, 659, 663 selection for applications, 553 activated carbon fibers, 40, 431 activation-pore structure relationship in, 444 advantages o£ 431 applications o£ 447 characterization o£ 436 preparation o£ 431 activated mesocarbon microbeads, 113 active surface area, 598 adsorption calorimetry, 57 adsorption energy distribution (AED), 9, 212, 339 adsorption energy surface (AES), 214 thermodynamic meaning o£ 149 adsorption enthalpy, 5, 53, 56, 336, 526, 588
adiabatic, 68 differential, 6, 226 integral, 68 isosteric, 56, 61, 293, 336 isothermal, 67 adsorption from solution, 273 for carbon characterization, 289 isotherm trypes, 291 thermodynamics o£ 290 adsorption hysteresis, 10, 459 adsorption isotherms, types, 7 adsorption potential, 147 distribution (APD), 462, 469 adsorption thermodynamics, 53 classical thermodynamics, 54 statistical mechanics, 59 adsorptive processes, design o£ 585 aging of carbons, 305 albumin adsorption, 358 algal metabolites/toxins adsorption, 696, 699 alkane adsorption, 521 alkanethiol adsorption, 522 alpha plot method, 9, 470 ammonia adsorption, 179 anion exchange properties, 318 argon adsorption, 80, 255, 337, 414 atomic force microscopy (AFM), 516 atrazine adsorption, 690 bacteria adsorption, 671 Barrett, Joyner and Halenda (BJH) method, 246, 461 basic structural unit (BSU), 25 basicity of 1T-electrons, 316 benzene adsorption, 550, 664 biomedical fluids adsorption, 716 bivariate model, 213, 225 bivariate surfaces, 213 boron doping, 504, 604 737
Subject Index
Broekhoff-de Boer method, 246 Brunauer-Emmett-Teller (BET) theory/ equation, 3, 473 C 6o ,329 hydrogenation o£ 348 C 7o ,329 calorimetry, 57 canonical ensemble, 60, 62, 93 capacitance, 607 carbon, 17 allotropes and polytypes, 17 alloys, 21 nanotexture, 16, 28, 38 phase diagram for, 20 structures, 17 carbon anodes, nanostructurated, 597, 602, 607 carbon black, 4, 5, 34, 35, 255, 460, 464 carbon dioxide adsorption, 88, 91, 179, 244, 331, 344, 438 carbon electrodes, biologically active, reactions at, 502 corrosion processes in, 503 chemically modified, 492 electrochemical kinetics on, 494 in molten salts, 504 manufacturing techniques for, 506 modified by transition metal complexes, 499 organic electrochemistry at, 501 oxygen electroreduction (OERR) on, 495 in acid solutions, 498 in alkaline solutions, 497 reactions at, 499 surface oxygen complexes in, 493 surface radical states in, 486 thermodynamics o£ 484 types of carbons for, 485, 486 carbon fibers, 23, 32, 33, 34 carbon membranes, for gas separation, 578 carbon molecular sieves, 7, 572 carbon monoxide adsorption, 343 carbon nanofibers, 32, 403, 406 carbon nanotubes, 16, 187, 369 activated, as electrode materials, 618 as anodes for Li-ion batteries, 600
as electrodes for supercapacitors, 616 functionalization, 617 mUltiwall (MWNT), 30, 407 single-wall (SWNT), 30, 187, 369 axial phase transition in, 194 bundles o£ 188, 369 charging with alkali metals, 373, 383 endohedral adsorption on, 190, 372, 376, 383 exohedral adsorption on, 202 endohedral transitions in, 196 interstitial sites/channels, adsorption on, 198,376 opening o£ 378 carbon surfaces, fractality o£ 490 nitrogen sites in, 322 roughness o£ 489 oxygen complexes in, 305 acidity distribution, 310 characterization, 306, 307 generation, 305, 306 cation exchange properties, 312 CD 4 adsorption, 413 classical thermodynamics, 54 CMK-l, 457, 467 CMK-3, 456, 459, 466, 471, 474 colloid imprinted carbons, 42 computer simulations of adsorption, 77 boundary conditions, selection of, 81 ensemble, selection o£ 82 generating configurations, 83 initialization, 83 potential energy surface, 79 conducting polymers, 619 contact angle, 168 cylindrical pore, 11, 245 grand canonical Monte Carlo (GCMC) simulation o£ 257, 280 dangling bonds, 301, 302 decWorination and decWoramination, 644 density functional theory (DFT), 10, 253 ab initio, 341, 375 nonlocal (NLDFT), 10, 64, 253 application to pore size distribution (PSD) determination, 253, 287 thermodynamic (TDFT), 375
Subject Index
deuterium adsorption, 390, 419 diamond, 17 disordered carbons, 605 doping with heteroatoms, 602 Dubinin theory, 7, 247 Dubinin-Radushkevich equation, 7, 247 characteristic curve, 7, 440 dye adsorption, 666 electrical double layer, 480, 487, 607 electrochemical energy storage, 593 electrochemical interface, 479 electrochemical kinetics, 482 electrodes, adsorption at, 481 electrolyte adsorption, 660 electrophoresis, 319 endotemplating and exotemplating, 456 energetic heterogeneity, 262 energetic topography, 211 and attractive interactions, 228 and repulsive interactions, 227 scaling behavior, 230 temperature dependence, 230 enhanced potential method, 250 ensemble and time averaging, 91 Escherichia coli adsorption, 672, 673 ethylene adsorption, 333 exfoliated graphite, 5, 29, 712 macropore structure in, 713 Frenkel-Halsey-Hill (FHH) equation, 6 fullerene, 35, 37, 329 as phase for cleaning and preconcentrating analytes, 356 defective fullerene, 333 hydrogenation of': 348 lattice hydrogen in, 346 porosity in, 330 water solutions of': 357 fullerene black, 35, 331 gas chromatography, 355 gas mixture adsorption, 59, 65, 69, 334 gas separation, 567 gas-solid adsorption, 3 energetics of': 53 gas-solid virial coefficients, 218 generalized Gaussian model (GGM), 213, 216 comparative test for, 223
739
geosmin adsorption, 696 Gibbs adsorption equation, 170, 480 Gibbs ensemble, 96, 258 glass-like carbon, 37 gold cyanides, adsorption of': 322, 323 grand canonical ensemble, 96 grand canonical Monte Carlo (GCMC), 10,96,121,124,226,257,280,332 graphene, 19, 42, 104, 515 curved graphene structures, 383 graphite, 18, 28, 80, 176 basal plane, 81, 180, 514 ion intercalation in, 490 structure, 18 graphitizable carbons, 23 heavy oil, 711 sorption, 716 factors affecting 71 7 kinetics, 722 recovery, 727 helium adsorption, 337, 417 high-performance liquid chromatography (HPLC), 353 highly oriented pyrolytic graphite (HOPG), 28, 514 superstructures in, 518 hormone adsorption, 695 Horvath-Kawazoe method, 248 humic substances adsorption, 669 hydrated transition metal ions, adsorption of': 322 hydrofullerene, 346 hydrogen adsorption, 346, 369, 419 and hydrogen storage, 346, 370, 403, 404 at cryogenic temperatures, 374 endohedral adsorption, 346, 350, 383 isotopes, 388 modeling of chemisorption, 384 modeling of phyisorption ab initio, 379 with classical potentials, 371 phase transitions, 391 axial, 194 production, from reformer off-gas, by pressure swing adsorption, 574 hydrogen cyanide adsorption, 545 hydrogen sulfide adsorption, 534, 646 adsorption-oxidation mechanism, 536
740 hydrophilic surface sites, 11, 302 hydrophobic carbon surfaces, 302 hysteresis loops, 7, 461 ideal adsorption solution theory (lAST), 70 ideal heterogeneous systems, simulations o£ 221 immersion calorimetry, 274 into pure liquids, 274 setup for nonwetting systems, 278 setup for wetting systems, 276 immersion enthalpy, 282, 663 immersion thermodynamics, 280 infrared spectroscopy of surface species, 343 inorganic gases, adsorption o£ 534 inorganic solutes, adsorption o£ 631 integral equation of adsorption, 151 resolution o£ 152 analytical solutions, 152 numerical solutions, 152 intercalation, 595 internal energy, 66 inverse gas chromatography, 338 iodine adsorption, 296 ionic strength, effect on adsorption, 670 irreversible capacity, 597 isoelectric point, 319, 321 Kelvin equation, 3, 10, 251 krypton adsorption, 409 lead adsorption, 356 Lennard-Jones equation/potential, 79, 108, 148, 213, 241 light oils sorption, 716 lithium insertion, 595 mechanism, 605 lithium-ion battery, 595 mass titration, 320 metal ion adsorption, 632 mechanism, 633 methane adsorption, 175, 205, 409, 412 and methane storage, 587 methyl tertiary-butyl ether (MTBE) adsorption, 693
Subject Index
microbial colonization, 671, 687 microcystins adsorption, 699 nucropore characterization, contribution of activated carbon fibers to, 438 filling, 4, 9 microporous carbons, as supercapacitor electrodes, 609 molecular models for porous carbons, 106 ab initio simulation methods, 119 reconstruction methods for, 107 reverse Monte Carlo (RMC) , 98, 110 constrained reverse Monte Carlo (CRMC), 114 regular porous carbons, 106 semiempirical methods, 119 simple geometric models, for disordered porous carbons, 107 monolayers, self-assembled, 521 molecular dynamics (MD), 83, 337 Monte Carlo method, 10, 85, 257, 471 Metropolis method, 86 grand canonical Monte Carlo (GCMC), 10, 96, 121, 124, 226, 257, 280, 332 naphtalene adsorption, 352 natural gas storage, 587 natural organic matter removal by adsorption, 668, 688, 702 neon adsorption, 203, 422 n-heptane adsorption, 338 nitrogen adsorption, 5, 243, 255, 331, 424, 438, 458 doping,603 production from air, by pressure swing adsorption, 572 nitrogen oxides adsorption, 343, 546 noble gas adsorption, 175, 337, 408 nonelectrolyte adsorption, 658 nongraphitizable carbons, 23, 37 nonporous carbons, physisorption on, 5, 7 multilayer isotherms in, 5 oil spills, remediation o£ 711 ordered mesoporous carbons, 41, 455 analysis o£ by XPS, 467
Subject Index
applications o£ 457 graphitic character of the surface o£ 465, 469 pore size distribution in, 461 ordered microporous carbons, 42 organic solutes, adsorption, 653 oxygen adsorption, 341 path integral Monte Carlo (PIMC) method, 98 pesticide adsorption, 690 pH of carbons, 319 phase transitions, 95 phenol adsorption, 660 mechanisms o£ 661, 663 physisorption, 3 point of zero charge, 320 polarity, 177 polycyclic aromatic hydrocarbons (PAHs) adsorption, 354 pore classification, IUPAC, 4 pore models, 103 pore size, 610 analysis, by adsorption from solution, 295 classification, 240 distribution (PSD), 6, 9, 12, 122, 443, 461 numerical inversion for determining, 262 regularization method for determining, 263 porous carbons confinement in, 125 nanotexture in, 38 porous texture, 239, 273 characterization by gas-solid adsorption, 239 characterization by immersion calorimetry, 286 characterization by liquid-solid adsorption, 273 potential models, 240 fluid-fluid, 241 and solid-fluid potential energy, 244 pressure swing adsorption process, 570, 572, 574, 576 pyrone-like structures, 314
74 1 quantum sieving, quantum molecular sieves, 385 reaction with aqueous bases, 308, 309 reverse Monte Carlo (RMC) method, 98 reversible capacity, origins o£ 595 scanning tunneling microscopy (STM) , 516 schwarzite, 39, 333 slit-shaped pores, 11, 104, 109, 173, 240, 244, 372 grand canonical Monte Carlo (GCMC) simulation o£ 257, 280 small-angle X-ray scattering (SAXS), 445 solvent vapor recovery, by adsorption 570 submonolayers, self-assembled, 521 sulfur atom submonolayers, 522 sulfur dioxide adsorption, 542 adsorption-oxidation mechanism, 542 supercapacitors, electrochemical, 607 superhydrophobicity, 302 surface area, 1, 12, 286, 295, 473 surface chemistry characterization by immersion calorimetry, 283 characterization by liquid-solid adsorption, 273 surface complexation models (SCM), 636 surface heterogeneity, 8, 147, 233 surface tension, 168, 171 surfactant adsorption, 666 taste and odor removal, from potable water, 696 templated carbons, 41, 457 as supercapacitor electrodes, 613 tetrafluoromethane adsorption, 422 thermal swing adsorption process, 570, 571 thiol adsorption, 341 tricholoroethylene (TCE) adsorption, 694 vinyl cWoride monomer, adsorption o£ 581 viscous organics sorption, 711 volatile organic compounds (VOCs), adsorption/removal, 549, 581
Subject Index
74 2 water adsorption, 11, 122, 176,583 water treatment, 631, 679 factors influencing, 681 wettability characterization, 284 wetting isotherm, 172 wetting of solids by liquids, 167
X-ray photoelectron spectroscopy (XPS),467 xenon adsorption, 411, 420 Young equation, 168 zeolite-templated carbons, 42 zeta potential, 319
AUTHOR INDEX
Arvia, AlejandroJ., 479, 513,19,20 Bandosz, TeresaJ., 533, 21 Beguin, Franvois, 593, 23 Bock, Henry, 103, 5 Boehm, Hans-Peter, 301, 13 Bojan, Mary J., 77, 187, 4, 9 Bolzan, Agustin E., 479,19 Bottani, Eduardo J., 53, 3 Calbi, M. Mercedes, 187, 9 Cazorla-Amor6s, Diego, 431, 17 Cole, Milton W., 187,369,9, 15 Darmstadt, Hans, 455, 18 Denoyel, Renaud, 273, 12 Do, Duong D., 239, 11 Do, Ha D., 239, 11 Faur-Brasquet, Catherine, 631, 24 Frackowiak, Elzbieta, 593, 23 Gatica, Silvina M., 187, 9 Gubbins, Keith E., 103, 5
Martinez-Alonso, Amelia, 329, 14 Migone, Aldo D., 403, 16 Moreno-Castilla, Carlos, 653, 25 Newcombe, Gayle, 679,26 Olivier, James P., 147,7 Pikunic, Jorge, 103, 5 Ramirez-Pastor, AntonioJ., 211,10 Riccardo, Jose L., 211, 10 Rouquerol, Franvoise, 273, 12 Rouquerol, Jean, 273, 12 Ryoo, Ryong, 455, 18 Salvarezza, Roberto C., 513, 20 Sing, Kenneth S.W., 3, 1 Sircar, Shivaji, 565, 22 Steele, William A., 77, 167, 4, 8 Suarez-Garcia, Fabian, 329, 14
Inagaki, Michio, 711, 27 Iwashita, Norio, 711, 27
Tasc6n, Juan M.D., 15, 53, 329, 2, 3, 14 Teran Arce, Fernando, 513, 20 Toyoda, Masahiro, 711, 27
Jakubov, Timur S., 133, 6 Johnson, J. Karl, 187, 369, 9, 15
Ustinov, Eugene A., 239, 11
Kang, Feiyu, 711, 27
Vela, Maria E., 513,20
Le Cloirec, Pierre, 631, 24 Linares-Solano, Angel, 431, 17
Zgrablich, Giorgio, 211, 10 Zubimendi, Jose L., 513, 20
735