A theoretical characterization of the interaction of water with oxidized carbonaceous clusters

A theoretical characterization of the interaction of water with oxidized carbonaceous clusters

CARBON 4 8 ( 2 0 1 0 ) 1 5 7 0 –1 5 7 9 available at www.sciencedirect.com journal homepage: www.elsevier.com/locate/carbon A theoretical characte...

743KB Sizes 0 Downloads 3 Views

CARBON

4 8 ( 2 0 1 0 ) 1 5 7 0 –1 5 7 9

available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/carbon

A theoretical characterization of the interaction of water with oxidized carbonaceous clusters M. Oubal a, S. Picaud a b

a,*

, M.T. Rayez b, J.C. Rayez

b

Institut UTINAM–UMR 6213, CNRS/Universite´ de Franche-Comte´, 16 Route de Gray, F-25030 Besanc¸on Cedex, France Institut des Sciences Mole´culaires – UMR 5255, CNRS/Universite´ de Bordeaux I, 351 Cours de la Libe´ration, F-33405 Talence Cedex, France

A R T I C L E I N F O

A B S T R A C T

Article history:

Quantum mechanical calculations are used to study the interaction of water molecules

Received 5 August 2009

with oxidized carbonaceous clusters, modeling soot primary particles emitted by aircraft

Accepted 22 December 2009

engines. First, the interaction of atomic oxygen with the face or the edges of small graphite

Available online 29 December 2009

crystallites of nanometer size is modeled at a quantum level, leading to the formation of graphene-oxide-like clusters. Then, a similar quantum approach is used to characterize the adsorption of one to six water molecules on the resulting clusters, in order to better understand, at a molecular level, the role of such oxidized sites on the hydrophilic properties of soot. It is shown that epoxide-like sites can participate in the hydrophilic behavior of soot primary particles although they are less attractive for water than hydroxyl or carboxyl sites. Ó 2009 Elsevier Ltd. All rights reserved.

1.

Introduction

Combustion-derived solid particles are suspected to have a non-negligible impact on the Earth’s radiative balance [1–5] by accounting for both direct heating due to light absorption and indirect cooling because of their activation as ice and cloud condensation nuclei [6,7]. In addition to natural aerosols coming, for example, from biomass burning [8], soot emitted by engines is currently one of the largest sources of uncertainties in understanding the fossil fuel burning impact on climate. This is mostly due to the high natural variability of combustion particle properties in respect to their sources, conditions of combustion and fuel composition [9]. In contrast to the other engine-soot produced at the ground level, aircraft-generated soot is directly emitted in the upper troposphere-lower stratosphere (UTLS) region where it may allow heterogeneous freezing of water and ice nucleation at much lower pressures than those required for homogeneous freezing [10]. This results in the formation of condensation trails behind planes that may evolve into artificial cirrus clouds

which add their effect on climate to natural cirrus clouds [11–13]. However, a quantitative study of the impact of soot on climate is a challenging task because it requires geometrical, chemical and optical characterizations of the primary carbonaceous particles constituting soot [14]. Widely accepted structural models for combustion soot assume the presence of graphene sheets consisting of a net of hexagons formed by carbon atoms as in graphite. The stacking of these graphene sheets on concentric spheres of different radii results in typical onion-like structures containing also other elements (mainly oxygen) besides carbon [15]. Recent transmission electron microscopy (TEM) studies have evidenced that soot emitted by aircraft is also made of nanocrystallites containing graphite-type layers arranged in an onion-like structure. Aircraft soot nanoparticles are thus of quasi-spherical shape, with diameters ranging between 10 and 50 nm [16]. Moreover, the graphite layers are partially oxidized, and they contain a certain number of oxygen atoms sites [17,18]. It is, however, worth noting that experimental characterizations of aircraft

* Corresponding author: Fax: +33 3 81 66 64 75. E-mail address: [email protected] (S. Picaud). 0008-6223/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.carbon.2009.12.054

CARBON

4 8 ( 20 1 0 ) 1 5 7 0–15 7 9

soot are quite scarce in the literature due to the difficulty of collecting real soot behind a plane. Measurements of water adsorption isotherms on laboratory-made soot and on aircraft engine combustor soot confirmed that soot can acquire a substantial amount of water molecules, and that soot from different sources demonstrates a wide variety of hydration properties [19,20]. In a recent series of papers, we used quantum calculations [21–23], molecular dynamics simulations [24–26] and Grand Canonical Monte Carlo (GCMC) simulations [27–29] to characterize, at a molecular level, the interaction of water with various carbonaceous nanoparticles modeling the primary particles constituting soot emitted by aircraft. Our results showed that adsorption of water on soot depends both on the primary particle structure (size of the pores) and on the type and location of the chemical defects that may be present in/on these primary particles. In particular, COOH sites on soot are much more attractive for water than OH sites, because of their larger ability to form hydrogen bonds with water molecules. To further increase our understanding on the ice nucleation properties of these primary particles of soot, we consider here a new class of carbonaceous particles resulting from a direct chemical attack by atomic oxygen. Indeed, atomic oxygen is one of the more reactive species in the atmosphere (especially the excited singlet state O(1D)) [30], and it can likely participate in the formation of oxidized soot particles that may differ from the particles considered in our previous studies and, as a consequence, may adsorb different amounts of water. As in our previous papers [21–23], we model the surface of a primary particle of soot by a C80H22 cluster and we make use of quantum calculations based on the ONIOM (Our own N-layered Integrated molecular Orbital and Molecular mechanics) method [31–33] to characterize the structure of the oxidized cluster and of small water aggregates adsorbed on it. It is worth noting that this study could also be of interest for other fields of applications related, for instance, to the characterization of graphene-oxide properties. Indeed, thin sheets of graphene-oxide have recently emerged as new carbon-based nanoscale materials, with solubility in water and other solvents that allow an uniform deposition onto wide ranging substrates making it potentially useful for macroelectronics [34]. However, despite these potentially important applications, theoretical studies on the oxidation of carbonaceous surfaces by atomic oxygen are relatively scarce [35–42], and among them, only a small number has been devoted to the oxidation of small graphene clusters [38,40,41]. Although these studies evidenced the formation of epoxide-like groups at the carbonaceous surface as a consequence of the attack by atomic oxygen, large differences have been found in the calculations of the corresponding adsorption energies depending on the quantum methods used. The present study based on the ONIOM approach can thus help at better characterizing this oxidation process. Moreover, as far as we know, this paper is the first one reporting quantum calculations on the interaction of water with graphene-oxide clusters. The paper is organized as follows: Section 2 gives a brief description of the technical part of our calculations. Section 3.1 presents the results obtained when considering the oxidation of the graphene surface by atomic oxygen, whereas in

1571

Section 3.2 the structure and energies of small water aggregates adsorbed on the corresponding oxidized cluster are discussed. In Section 3.3 the same approach is used to characterize the oxidation process at the edge of a large carbonaceous cluster, and the main conclusions of the present study are given in Section 4.

2.

Computational details

The method used in the present work is similar to the one developed in our previous papers [21–23]. The graphene-like nanoparticle is modeled by 30 fused benzene rings arranged in a single atomic layer containing 80 carbon atoms, the edges of which being saturated by hydrogen atoms (Fig. 1a). Hereafter, this cluster will be referred as the C80H22 cluster. First, the ONIOM two-layer method [31–33] is used to model the oxidation of the C80H22 cluster by atomic oxygen. Then it is used to characterize the interaction of small water aggregates (from one to six water molecules) with the oxidized cluster. Using the ONIOM method, the central part of the system that contains the adsorbate (atomic oxygen or water molecules) and the closest neighboring C atoms (typically the closest four carbon rings, although some tests have been performed with another partition including 14 carbon rings, see Fig. 1) is treated with a high-level of accuracy by using density functional theory (DFT) whereas the rest of the system is taken into account with the semi-empirical PM3 (Parameterized Model No. 3) method [43,44]. On the basis of our previous works [21–23], the high-level DFT calculations are performed using the B3LYP (Becke 3-Parameter, Lee, Yang and Parr) exchange–correlation functional [45] and cc-PVDZ (Correlation-Consistent

Fig. 1 – Geometry of the C80H22 clusters used in our ONIOM calculations based on two different partitions between the DFT and the PM3 layers: the C atoms treated at the highlevel of quantum calculations in the ONIOM method are represented as small balls whereas the other atoms are represented by sticks; (a) partition A, (b) partition B and (c) view of the bridge, top and hexagonal sites above the center of the C80H22 cluster. Note that we have also indicated on (a) the edge site considered in our calculations for the edgeoxidation of the C80H22 cluster (see text).

1572

CARBON

4 8 ( 2 0 1 0 ) 1 5 7 0 –1 5 7 9

Polarized Valence Double-Zeta) basis set [46]. All the calculations have been carried out by using the Gaussian 03 quantum chemistry package [47] and the stable geometries have been calculated with a tight criterium geometry optimization. Note that once the geometry of the adsorbed water aggregates is optimized, the final energy is calculated with the larger 6311++G(2d,2p) basis set (which is a triple-zeta split valence set augmented with diffuse and polarization functions on all atoms) for the DFT layer, to minimize the basis set superposition errors (BSSE).

3.

Results and discussion

3.1. Interaction of atomic oxygen with the face of the C80H22 cluster

curves are shown in Fig. 2. They show that the energy difference above the three sites considered here is very small at large separations from the surface, but it strongly increases when approaching the C80H22 cluster, with a strong preference for the bridge site. These results are very similar to those recently published by Incze et al. [37]. Above the most stable site (i.e., the bridge site), the adsorption of atomic oxygen on the face of the C80H22 cluster gives rise to a highly stable epoxide-like structure (Fig. 3a), consistent with experimental findings [48] and previous calculations [35–37,39] on graphite surfaces. The calculated binding energy for the oxygen atom is equal to 98.5 kcal/mol (96.0 kcal/mol taking into account the correction coming from the zero-point vibrational energy). Note that we have defined the binding energy as DEb ¼ E½C80 H22 O  E½C80 H22   E½O;

The first stage of our study is devoted to the oxidation by atomic oxygen of the C80H22 cluster, aiming at modeling the oxidation process of soot particles behind planes. In the troposphere, O atoms are mainly produced by the photolysis of the oxygen and ozone molecules [30], leading to the formation of the oxygen atoms in both excited singlet O(1D) and ground triplet O(3P) states. Because O(1D) is likely more reactive in the atmosphere than O(3P) [30] we consider here the oxidation of the C80H22 cluster by one oxygen atom in the first 1D excited state only. Note that this is supported by recent quantum calculations which have shown that the oxidation of graphite and carbon nanotubes is much more favored by O(1D) than by O(3P) [36]. We thus investigate the approach of a O(1D) atom to the cluster surface over three high symmetry sites located in the middle of the cluster surface. These three sites are above the center of a hexagon, on top of a C atom and on a bridge site (Fig. 1b). The ONIOM calculations are performed using restricted DFT (B3LYP functional with cc-PVDZ basis set) for several z distances with respect to the surface of the cluster, the O(1D) atom being constrained to stay above a specific position whereas the other degrees of freedom of the whole system (i.e., C80H22 + O) are optimized. The resulting potential energy

Fig. 2 – Minimum energies calculated for different distances between the O(1D) atom and the C80H22 cluster, when considering approach above the three different sites of the cluster surface.

ð1Þ

where E[O] is the energy of the isolated oxygen atom calculated at the DFT level, E[C80H22] is the total energy of the optimized C80H22 cluster in the absence of the oxygen atom, and E[C80H22O] is the energy of the oxidized cluster, these two later energies being calculated using our ONIOM strategy. This binding energy thus accounts not only for the bond dissociation energy but also for the energy difference coming from all possible structural changes (i.e., deformation) of the carbonaceous cluster upon oxidation. Indeed, in the corresponding optimized structure, the two carbon atoms involved in the epoxide-like group need to undergo configuration changes from a planar sp2 hybridized to a distorted sp3-hybridized geometry which leads to a loss of planarity of the C80H22 cluster (Fig. 3a, side view). Moreover, these two C atoms are slightly pulled out of the cluster surface. The optimized C–O ˚ , and the epoxide C–C bond in the distance is equal to 1.43 A ˚ compared to the three-membered ring is elongated to 1.54 A ˚ value of the C@C bond in the non-oxidized cluster it1.42 A self. These results are similar to those found on a perfect graphite surface [35–37,39]. To assess the accuracy of the DFT calculations, we have also considered different density functionals and basis sets

Fig. 3 – Optimized geometry (top and side views) of the optimized (a) epoxide and (b) open ring (oxepine) structures. Black, and grey circles represent O and C atoms included in the DFT cluster. The rest of the system is represented by grey (C) and white (H) sticks.

CARBON

4 8 ( 20 1 0 ) 1 5 7 0–15 7 9

Table 1 – Binding energy (in kcal/mol) of the O(1D) atom at the bridge of the carbonaceous surface calculated when using different density functionals and different basis sets to describe the first-layer of the ONIOM method. Basis set cc-PVDZ 6-311G(2d,2p)

B3LYP[45] 98.5 99.1

mPW1LYP[49] 95.8 96.6

B3PW91[50] 105.1 106.6

to calculate the binding energy of the O(1D) atom at the epoxide-like site. The corresponding results, given in Table 1, show that changing the functional and/or the basis set modifies the O binding energy by no more than 7–8%. Also, the calculations performed by including a larger number of C rings in the DFT (B3LYP – cc-PVDZ) layer (partition B, Fig. 1b) lead to similar geometrical results (although the deformation of the cluster is slightly less pronounced) and to a similar value of the corresponding adsorption energy (99.8 kcal/mol). However, these calculations are much more time consuming due to the larger number of atoms included in the DFT calculations, and only the smallest DFT layer (partition A, Fig. 1a) will be considered in the calculations when characterizing the adsorption of small water aggregates on the oxidized cluster. We have also performed additional calculations with unrestricted DFT to account for the possible formation of final triplet-biradical products above the top and hexagonal sites. These calculations show that such products are however less stable than the singlet states found above (Fig. 2), and much less stable than the singlet epoxide found above the bridge site.

1573

Finally, note that another possible structure is also found in the optimization procedure with restricted DFT (B3LYP functional with cc-PVDZ basis set) corresponding to the formation of an oxepine (Fig. 3b), i.e., to the breaking of the C– C bond. This structure is however less stable by about 2.2 kcal/mol and more distorted than the epoxide-like structure, in agreement with recently published results on graphite oxide surfaces [40,41]. Note that this small stabilization energy difference between the epoxide-like and the oxepine-like structures should facilitate an interconversion process between these two forms.

3.2. Adsorption of water molecules on the oxidized C80H22O cluster To characterize the hydrophilicity of the C80H22O grapheneoxide cluster optimized above, we investigate here the details of the adsorption process of small water aggregates containing up to six water molecules. These water molecules together with the four central carbon rings and the epoxidelike group are treated with a high-level of accuracy (i.e., they are included in the first-layer of the ONIOM method) whereas all the other atoms are taken into account at the semi-empirical level. Note that the initial positions and orientations (i.e., before optimization) of the water molecules have been chosen by informed guesses, assuming that the optimized configuration would correspond to the formation of hydrogen bonds between the water molecules and the carbonaceous cluster as well as within the water aggregate. The most stable geometry for one single water molecule adsorbed on the cluster is given in Fig. 4a. It exhibits the formation of one hydrogen bond between the O atom of the

Fig. 4 – Optimized geometry for the stable structure of small water aggregates (H2O)n adsorbed on the C80H22O cluster. (a) n = 1, (b) n = 2, (c) n = 3, (d) n = 4, (e) n = 5, (f) n = 6 water molecules. Black, grey, and white circles represent O, C, and H atoms included in the DFT cluster. The rest of the system is represented by grey (C) and white (H) sticks.

1574

CARBON

4 8 ( 2 0 1 0 ) 1 5 7 0 –1 5 7 9

Table 2 – Mean adsorption energy per molecule (DEads/n) and the corresponding incremental (i.e., stabilization) energy DEinc above the face of the defective cluster C80H22O for small water aggregates containing up to n = 6 water molecules. These energies are given in kcal/mol and they include the (weak) counterpoise corrections to the basis set superposition error (BSSE corrections). n (H2O) 1 2 3 4 5 6

DEads/n

DEinc

2.89 3.61 5.97 7.10 6.98 6.76

2.89 4.33 10.70 10.49 6.50 5.67

surface and one hydrogen atom of the water molecule, in which the water molecule acts as a proton donor. However, because the corresponding adsorption energy value is equal to 2.89 kcal/mol, only, with the OH distance being equal to ˚ and the OHO angle equal to 163°, this hydrogen bond 2.03 A appears quite weak with respect to optimized hydrogen bonds between, i.e., water molecules in ice. Moreover, the adsorption process is accompanied by a very small charge transfer of 0.01e from the cluster to the water molecule, indicating a physisorption process. Note that additional calculations were also performed with another basis set d95V++G(2d,2p) and another density functional MPW1LYP to assess the accuracy of these results. The corresponding results are similar to those obtained above (between 2.89 kcal/mol when using B3LYP with d95V++G(2d,2P) basis set and 3.5 kcal/mol when using MPW1LYP with d95V++G(2d,2P) basis set), and we thus decided to perform the calculations for larger water aggregates with the B3LYP functional and the 6-311G++(2d,2p) basis set, for the comparison with our previous works [21–23]. The mean adsorption energies per water molecule (DEads =n) obtained when increasing the number n of adsorbed water molecules above the C80H22O cluster are given in Table 2, and the corresponding stable geometries are given in Fig. 4. The adsorption energies for aggregate containing n water molecules are determined by: DEads ¼ E½C80 H22 O  ðH2 OÞn   E½C80 H22 O  nE½H2 O

ð2Þ

where E[H2O] is the energy of the isolated water molecule optimized at the DFT level, E[C80H22O] is the total energy of the relaxed C80H22O cluster in the absence of the adsorbed water molecule, and E[C80H22O  (H2O)n] is the energy of the adsorbed system, these two later energies being calculated using our ONIOM strategy. The mean adsorption energy DEads =n first decreases with n, from 2.89 kcal/mol (n = 1) down to about 7 kcal/mol for an aggregate containing four water molecules. Then, an almost constant value around 7 kcal/mol is calculated when considering one and two extra adsorbed molecules, indicating the saturation of the hydrophilic epoxide-like site. Indeed, for large water aggregates, the equilibrium configurations (Fig. 4) and the adsorption energies are governed by the lateral interactions between water molecules rather than by the interactions between the water molecules and the surface.

Fig. 5 – Adsorption energy per molecule (DEads/n) calculated for small water aggregates adsorbed on various oxidized C80H22 clusters containing epoxide, hydroxyl or carboxyl groups anchored on their face. Another way to characterize this saturation is to calculate the incremental association energy (i.e., the stabilization energy) defined as DEinc ¼ E½C80 H22 O  ðH2 OÞn   E½H2 O  E½C80 H22 O  ðH2 OÞn1 ;

ð3Þ

and corresponding to the energy lost by the system upon the addition of one water molecule. Table 2 shows that the maximum energy loss is obtained by adding a third molecule on the epoxide-like site, a situation that corresponds to the optimization of the hydrogen bonds network between the water molecules and the C80H22O cluster (Fig. 4c). Indeed, the molecules tend to form an optimized water trimer tied to the surface by a donor hydrogen bond. The energy loss is however similar when considering the addition of a fourth water molecule, with the formation of an optimized water tetramer parallel to the surface of the cluster. It is also interesting to note that the attachment of the water aggregate to the epoxide-like site depends on the number of water molecules. Indeed, for an odd number of water molecules (n = 1, 3, and 5), the water aggregate is tied to the epoxide group by a single water molecule which forms a donor hydrogen bond with the surface (Fig. 4a, c, and e), whereas for an even number of water molecules (n = 2, 4, and 6), the water aggregate is tied to the epoxide group by two water molecules (Fig. 4b, d, and f). To compare the results obtained on a C80H22O cluster with those of previous quantum calculations performed on other oxidized carbonaceous clusters containing OH and COOH groups anchored at their surface [22,23], the adsorption energies of small water clusters adsorbed on these oxidized C80H22 clusters are shown in Fig. 5. Note that for this comparison, we have added here the optimization results for water aggregates containing six water molecules. Indeed, previously published results were limited to aggregates containing at most five water molecules. Fig. 5 clearly evidences that the mean adsorption energy per water molecule for small aggregates (n 6 4) is quite similar on OH and epoxide groups with however a slightly lower

CARBON

4 8 ( 20 1 0 ) 1 5 7 0–15 7 9

1575

Fig. 6 – Optimized geometry for the stable structure of small water aggregates (H2O)n adsorbed at the edge of the C80H22O cluster (top and side views). Black, grey, and white circles represent O, C, and H atoms included in the DFT cluster. The rest of the system is represented by grey (C) and white (H) sticks. Only a small part of the system around the epoxide group is shown, for clarity. energy on cluster containing hydroxyl rather than epoxide groups. In contrast, much lower adsorption energies are calculated for small water aggregates on cluster containing COOH group irrespective of the number of adsorbed water molecules (when n 6 4), indicating that COOH may have a stronger influence on the water nucleation process than the other chemical groups, in accordance with previous molecular dynamics and Grand Canonical Monte Carlo simulation results [25,26,28]. Note that for larger water aggregates (n = 5 and 6) the adsorption energy is mainly governed by the optimization of the lateral interactions between water molecules and, as a consequence, it is similar on clusters containing COOH and OH groups. However, the situation is somewhat different on the epoxide group, because the distances between the water molecules and the underlying C atoms of the cluster are smaller than on OH and COOH, resulting in a

Table 3 – Mean adsorption energy per molecule (DEads/n) above the edge of the C80H22O epoxide-like cluster and of the defective cluster C80H22O2 for small water aggregates containing up to n = 6 water molecules. These energies are given in kcal/mol and they include the (weak) counterpoise corrections to the basis set superposition error (BSSE corrections). n (H2O) 1 2 3 4 5 6

C80H22O

C80H22O2

4.40 5.57 6.12 6.40 6.93 7.46

5.45 6.36 7.03 7.15 7.72 8.72

Fig. 7 – (a) Optimized geometry of the C80H20O cluster corresponding to the oxidation at an edge site of the carbonaceous cluster considered here; note that the C atoms treated at the high-level of quantum calculations in the ONIOM method are represented as small balls whereas the other atoms are represented by sticks; (b) optimized geometry of the edge-oxidized cluster after the chemisorption and dissociation of one water molecule. Black, grey, and white circles represent O, C, and H atoms included in the DFT cluster. The rest of the system is represented by grey (C) and white (H) sticks.

1576

CARBON

4 8 ( 2 0 1 0 ) 1 5 7 0 –1 5 7 9

Fig. 8 – Optimized geometry for the stable structure of small water aggregates (H2O)n adsorbed at the edge of the C80H22O2 cluster (top and side views). Black, grey, and white circles represent O, C, and H atoms included in the DFT cluster. The rest of the system is represented by grey (C) and white (H) sticks. Only a small part of the system around the OH groups (see text) is shown, for clarity.

larger repulsive contribution of the carbonaceous surface to the total adsorption energy. In a general way, this small steric repulsion of the p-conjugated system explains why the water adsorption energies calculated on the graphene-oxide cluster is higher than on the other oxidized carbonaceous clusters containing OH or COOH groups. Note, however, that when considering large water aggregates (see Fig. 4f), this latter result may also depend on the size of the DFT layer because some water molecules are located near the boundary between the DFT and PM3 layers.

3.3.

Adsorption at the edge of the carbonaceous cluster

3.3.1.

Adsorption at the edge of a perfect cluster

After considering oxidation on the face of the C80H22 cluster, we also investigate the adsorption process at the edges of the carbonaceous cluster. We only consider oxidation by atomic oxygen O(1D) at the most symmetric site of the C80H22 cluster edges (Fig. 1) which allows the definition of a symmetric DFT cluster in our ONIOM approach (see Fig. 6). The most stable structure after oxidation corresponds to the formation of an epoxide-like group on the edge of the cluster (see Fig. 6), with an adsorption energy for the oxygen atom

equal to 135.0 kcal/mol, indicating that the oxidation at the edge is strongly favored with respect to the oxidation on the face of the cluster. Then, we investigate the adsorption process for small water aggregates (up to six water molecules) on this edge-oxidized cluster. The corresponding adsorption energies are given in Table 3 and the most stable geometries are shown in Fig. 6. In a general way, the water molecules tend to adsorb on one side of the cluster, the corresponding aggregate being slightly stabilized by the interactions with the hydrogen atoms that are located at the edges of the C80H22 cluster. This leads to adsorption energies that are slightly lower than on the face of the cluster, i.e., the water adsorption is favored at the edge with respect to the face of the cluster. Such a favored adsorption on edge could also come from a small steric repulsion effect of the p-conjugated system on the face of the cluster. It is also interesting to note that for large water aggregates containing more than four molecules, the adsorption energy mainly comes from the optimization of the hydrogen bond network between water molecules, and, as a consequence, the differences with adsorption on the face of the cluster are smaller than for small water aggregates. The comparison of these results with those previously obtained on graphene clusters containing OH and COOH groups

CARBON

4 8 ( 20 1 0 ) 1 5 7 0–15 7 9

at their edges [22,23] shows that the adsorption energies on epoxide group are higher than those calculated on OH and COOH groups, the water adsorption being thus much more favored on cluster containing COOH group than on the two other types of clusters. Moreover, for small water aggregates (typically n 6 3), the adsorption energies are closer on OH and epoxide groups located at the edge of the C80H22 cluster (for which the water aggregates form on one side of the cluster) than on OH and epoxide groups located on the face of the cluster (for which the water aggregates form on top of the surface). This result confirms the small repulsive effect of the underlying C atoms when considering adsorption on the face of the cluster (see above).

3.3.2.

Adsorption at the edge of a defective cluster

We also characterize here the oxidation on a defective cluster, i.e., after removing two nearest neighbor hydrogen atoms at the edge of the C80H22 cluster (again by considering the most symmetric position (see Fig. 7) only). The geometry of the resulting C80H20O system is optimized by using the ONIOM approach, in which the four nearest carbon rings are included in the DFT calculations (see Fig. 7). The optimized structure of the oxidized defective cluster corresponds to the formation of a very stable carbonyl group (Fig. 7a) which leaves the other carbon of the edge site unsaturated. The subsequent approach of one water molecule leads to its dissociation with the formation of two adjacent OH groups defining a C80H22O2 cluster (Fig. 7b). Although such water dissociation may be unlikely depending on its energy barrier in tropospheric conditions, the final structure corresponds to a possible strong trapping site for additional water. We thus characterize the adsorption of two to six water molecules on the corresponding C80H22O2 cluster. The corresponding mean adsorption energies are given in Table 3 and the stable structures for the water aggregates are shown in Fig. 8. The mean adsorption energies at the edge-oxidized defective cluster (Table 3) are always lower than on the face and at the edge of the oxidized perfect cluster, indicating that the water adsorption is favored on the defective cluster with respect to the adsorption on a perfect cluster. For n = 1, this can be easily explained by the possible formation of two hydrogen bonds between the water molecule and the two OH groups located at the edge of the cluster (top of Fig. 8), whereas only one hydrogen bond has been found above the epoxide group located both at the face and at the edge of the perfect cluster. However, the distance between these two OH groups is too large to optimize two strong hydrogen bonds with the adsorbed water molecule as previously obtained above a COOH group [21,23], and the corresponding site thus appears much less attractive than a COOH site. The formation of two weak hydrogen bonds between the carbonaceous cluster and larger water aggregates is also obtained for n P 2 although for these aggregates the main contribution to the adsorption energy comes from the optimization of the hydrogen bond network between water molecules. Note finally, that for n = 6, the calculations evidence a partial transfer of one hydrogen atom from on OH of the surface to the water aggregate, resulting in a lower adsorption energy.

4.

1577

Discussions and conclusions

The ONIOM two-layer method has been used to investigate the adsorption of small water aggregates above the face and at the edge of a large carbonaceous cluster oxidized by atomic oxygen. On the face and at the edge of this cluster, the oxidation leads to the formation of a stable epoxide-like site which behaves as a nucleation center for water, slightly more attractive when it is located at the edge rather than on the face of the cluster. However, the mean adsorption energy per water molecule shows that the affinity of such epoxidelike site for water is slightly weaker than that of a OH site, and much weaker than the affinity between COOH sites and water. When the oxidation process takes place at the defective edge of the carbonaceous cluster, an unsaturated site is created which could dissociate a first water molecule, leading to the formation of a double OH site. If existing in real soot, such double OH site is more attractive for water than the epoxide-like sites considered at the face and edge of the non-defective carbonaceous cluster, but it is less attractive for water than COOH sites considered in previous studies [21,23]. Nevertheless, the epoxide-like sites (on the face and at the edge of the carbonaceous cluster) and the double OH site can participate in the hydrophilic behavior of soot primary particles, and their presence should be considered in further modeling of the ice nucleation properties of soot. Moreover, soot being a very complex system, oxygen containing groups located not only on the face and on the edges but also between graphene sheets [51] should also be considered in future studies. Note that the present calculations are based on the popular hybrid functional B3LYP which might lead to questionable quantitative values for the adsorption energies because it is not the best functional to describe hydrogen bonded systems [52]. However, we make use of a large basis set including diffuse functions for the energy calculations, and this is known to give acceptable values for the energies of water cluster of small size when compared to high-level quantum calculations [52]. Such high-level quantum calculations are however not realistic here due to the large size of the systems considered. Moreover, the main interest of our work relies on the comparative characterization of the hydrophilic behavior of different chemical groups anchored on carbonaceous clusters and not on the calculations of very accurate adsorption energies. We thus believe that the main conclusions of this work do not strongly depend on the level of accuracy of the quantum calculations performed. Finally, it is worth noting that the present results can also be of interest for people working on graphene oxides properties.

Acknowledgements M. Oubal thanks the CNRS and the Re´gion de FrancheComte´ for his PhD grant. Dr. Martial Boggio-Pasqua is also gratefully acknowledged for helpful discussions on the ONIOM method.

1578

CARBON

4 8 ( 2 0 1 0 ) 1 5 7 0 –1 5 7 9

R E F E R E N C E S

[1] Andreae MO. The dark side of aerosols. Nature 2001;409:671–2. [2] Chung SH, Seinfeld JH. Global distribution and climate forcing of carbonaceous aerosols. J Geophys Res 2002;107:4407–40. [3] Lohmann U, Feichter J. Global indirect aerosol effects: a review. Atmos Chem Phys 2005;5:715–37. [4] Roeckner E, Stier P, Feichter J, Kloster S, Esch M, Fischer-Bruns I. Impact of carbonaceous aerosol emissions on regional climate change. Climate Dyn 2006;27:553–71. [5] Streets DG. Dissecting future aerosol emissions: warming tendencies and mitigation opportunities. Climatic Change 2007;81:313–30. [6] DeMott PJ, Chen Y, Kreidenweis SM, Rodgers DC, Sherman DE. Ice formation by black carbon particles. Geophys Res Lett 1999;26:2429–32. [7] Penner JE, Lister DH, Griggs DJ, Dokken DJ, McFarland M, editors. Intergovernmental panel on climate change (IPCC) special report: aviation and the global atmosphere. Cambridge: Cambridge University Press; 1999. [8] Andreae MO, Merlet P. Emission of trace gases and aerosols from biomass burning. Global Biogeochem Cycles 2001;15:955–66. [9] Chen Y, Penner JE. Uncertainty analysis for estimates of the first indirect aerosol effect. Atmos Chem Phys 2005;5:2935–48. [10] Cantrell W, Heymsfield A. Production of ice in tropospheric clouds: a review. Bull Am Meteorit Soc 2005;86:795–807. [11] Ka¨rcher B. Aviation-produced aerosols and contrails. Surv Geophys 1999;20:113–67. [12] Hendricks J, Ka¨rcher B, Do¨pelheuer A, Feichter J, Lohmann U, Baumgardner D. Simulating the global atmospheric black carbon cycle: a revisit to the contribution of aircraft emissions. Atmos Chem Phys 2004;4:2521–41. [13] Hendricks J, Ka¨rcher B, Lohmann U, Ponater M. Do aircraft black carbon emissions affect cirrus clouds on the global scale? Geophys Res Lett 2005;32:L12814:1–4. [14] Popovicheva O, Starik A. Aircraft-generated soot aerosols: physicochemical properties and effects of emission into the atmosphere. Atmos Ocean Phys 2007;43:147–64. [15] Kova´cs KV, Po´sfai M, La´ba´r JL. Nanostructure of atmospheric soot particles. Atmos Environ 2006;40:5533–42. [16] Popovicheva OB, Persiantseva NM, Trukhin ME, Rulev GB, Shonija NK, Buriko YY, et al. Experimental characterization of aircraft combustor soot: microstructure, surface area, porosity and water adsorption. Phys Chem Chem Phys 2000;2:4421–6. [17] Popovicheva OB, Persiantseva NM, Kuznetsov BV, Rakhmanova TA, Shonija NK, Suzanne J, et al. Microstructure and water absorbability of aircraft combustor soot and kerosene flame soot: toward an aircraftgenerated soot laboratory surrogate. J Phys Chem A 2003;107:10046–54. [18] Demirdjian B, Ferry D, Suzanne J, Popovicheva OB, Persiantseva NM, Shonija NK. Heterogeneities in the microstructure and composition of aircraft engine combustor soot: impact on the water uptake. J Atmos Chem 2007;56:83–103. [19] Alcala-Jornod C, Van den Bergh H, Rossi MJ. Can soot particles emitted by airplane exhaust contribute to the formation of aviation contrails and cirrus clouds? Geophys Res Lett 2002;29(1820):1–4. [20] Popovicheva O, Persiantseva NM, Shonija NK, DeMott PJ, Koehler K, Petters M, et al. Water interaction with hydrophobic and hydrophilic soot particles. Phys Chem Chem Phys 2008;10:2332–44.

[21] Hamad S, Mejias JA, Picaud S, Hoang PNM. A theoretical study of the adsorption of water on a model soot surface. I. Quantum chemical calculations. J Phys Chem B 2004;108:5405–9. [22] Collignon B, Hoang PNM, Picaud S, Rayez JC. Ab initio study of the water adsorption on hydroxylated surfaces. Chem Phys Lett 2005;406:431–6. [23] Collignon B, Hoang PNM, Picaud S, Rayez JC. Clustering of water molecules on model soot particles: an ab initio study. Comp Lett 2005;1:277–87. [24] Picaud S, Hoang PNM, Hamad S, Mejias JA, Lago S. A theoretical study of the adsorption of water on a model soot surface. II. Molecular dynamics simulations. J Phys Chem B 2004;108:5410–5. [25] Picaud S, Collignon B, Hoang PNM, Rayez JC. A molecular dynamics simulation study of the water adsorption on hydroxylated graphite surface. J Phys Chem B 2006;110:8398–408. [26] Picaud S, Collignon B, Hoang PNM, Rayez JC. Adsorption of water molecules on oxidized graphite surfaces: a molecular dynamics study of the competition between OH and COOH sites. Phys Chem Chem Phys 2008;10:6998–7009. [27] Moulin F, Picaud S, Hoang PNM, Partay L, Jedlovszky P. A Grand Canonical Monte Carlo simulation study of water adsorption on a model soot particle. Mol Simul 2006;32:487–93. [28] Moulin F, Picaud S, Hoang PNM, Jedlovszky P. Grand Canonical Monte Carlo simulation of the adsorption isotherms of water molecules on model soot particles. J Chem Phys 2007;127(164719):1–11. [29] Moulin F, Picaud S, Hoang PNM, Partay L, Jedlovszky P. A Grand Canonical Monte Carlo simulation study of the aggregation of water molecules on chemically modified soot particles. Comp Lett 2008;4:105–16. [30] Seinfeld JH, Pandis SN. Atmospheric chemistry and physics. New York: Wiley; 1998. [31] Svensson M, Humbel S, Morokuma K. Energetics using the single point IMOMO (integrated molecular orbital + molecular orbitals) calculations: choices of computational levels and model system. J Chem Phys 1996;105:3654–61. [32] Svensson M, Humbel S, Froese RDJ, Matsubara T, Sieber S, Morokuma K. ONIOM: a multilayered integrated MO + MM method for geometry optimizations and single point energy predictions. J Phys Chem 1996;100:19357–63. [33] Dapprich S, Koma´romi I, Byun KS, Morokuma K, Frisch MJ. A new ONIOM implementation in Gaussian98. Part I. The calculation of energies, gradients, vibrational frequencies and electric field derivatives. J Mol Struct (Theochem) 1999;461–462:1–21. [34] Mkhoyan KA, Contryman AW, Silcox J, Stewart DA, Eda G, Mattevi C, et al. Atomic and electronic structure of graphene-oxide. Nano Lett 2009;9:1058–63. [35] Lamoen D, Persson BNJ. Adsorption of potassium and oxygen on graphite: a theoretical study. J Chem Phys 1998;108:3332–41. [36] Sorescu DC, Jordan KD, Avouris P. Theoretical study of oxygen adsorption on graphite and the (8, 0) single-walled carbon nanotube. J Phys Chem B 2001;105:11227–32. [37] Incze A, Pasturel A, Chatillon C. Oxidation of graphite by atomic oxygen: a first-principles approach. Surf Sci 2003;537:55–63. [38] Ferro Y, Allouche A, Merinelli F, Brosset C. Theoretical study of oxygen adsorption on boron-doped graphite. Surf Sci 2004;559:158–68. [39] Jelea A, Marinelli F, Ferro Y, Allouche A, Brosset C. Quantum study of hydrogen–oxygen–graphite interactions. Carbon 2004;42:3189–98.

CARBON

4 8 ( 20 1 0 ) 1 5 7 0–15 7 9

[40] Li JL, Kudin KN, McAllister MJ, Prud’homme RK, Aksay IA, Car R. Oxygen-driven unzipping of graphitic materials. Phys Rev Lett 2006;96:176101. [41] Schniepp HC, Li JL, McAllister MJ, Sai H, Herrera-Alonso M, Adamson DH, et al. Functionalized single graphene sheets derived from splitting graphite oxide. J Phys Chem B 2006;110:8535–9. [42] Bergeron H, Rougeau N, Sidis V, Sizun M, Teillet-Billy D, Aguillon F. OH formation from O and H atoms physisorbed on a graphitic surface through the Langmuir–Hinshelwood mechanism: a quasi-classical approach. J Phys Chem A 2008;112:11921–30. [43] Stewart JJP. Optimization of parameters for semiempirical methods. I. Method. J Comput Chem 1989;10:209–20. [44] Stewart JJP. Optimization of parameters for semiempirical methods. II. Applications. J Comput Chem 1989;10:221–64. [45] Becke AD. Density-functional thermochemistry. III. The role of exact exchange. J Chem Phys 1993;98:5648–52. [46] Woon DE, Dunning Jr TH. Gaussian basis sets for use in correlated molecular calculations. III. The atoms aluminum through argon. J Chem Phys 1993;98:1358–71.

1579

[47] Frisch MJ, Trucks GW, Schlegel HB, Scuseria GE, Robb MA, Cheeseman JR, et al. Gaussian 03, revision C02. Wallingford, CT: Gaussian Inc.; 2004. [48] He H, Klinowski J, Forster M, Lerf A. A new structural model for graphite oxide. Chem Phys Lett 1998;287:53–6. [49] Adamo C, Barone V. Exchange functionals with improved long-range behavior and adiabatic connection methods without adjustable parameters: the mPW and mPW1PW models. J Chem Phys 1998;108:664–75. [50] Perdew JP, Chevary JA, Vosko SH, Jackson KA, Pederson MR, Singh DJ, et al. Atoms, molecule, solids, and surfaces: applications of the generalized gradient approximation for exchange and correlation. Phys Rev B 1992;46:6671–87. [51] Sergides CA, Jassim JA, Chughtai AR, Smith DM. The structure of hexane soot. Part III: Ozonation studies. Appl Spectrosc 1987;41:482–92. [52] Su JT, Xu X, Goddard III WA. Accurate energies and structures for large water clusters using the X3LYP hybrid density functional. J Phys Chem A 2004;108:10518–26.