A theoretical investigation of the removal of methylated arsenic pollutants with silicon doped graphene

Abstract

Quantum chemistry calculations were performed to study the adsorption of methylated arsenic pollutants onto silicon doped graphene (SiG), characterizing their geometrical parameters, energetic and binding properties. All the methylarsenicals are chemisorbed onto SiG by silicon–oxygen binding. Trivalent methylarsenicals are chemisorbed with adsorption energies of up to 1.1 eV at neutral conditions; while, pentavalent methylarsenicals reach energies of up to 1.8 and 4.0 eV in their neutral and anionic states, respectively. Additionally, it was determined that the pollutant stability after the chemisorption plays a key role in the adsorption strength; contributions from long-range interactions and charge transfer processes were also discussed. Finally, molecular dynamics and explicit solvent simulations indicated that the adsorption stability remains stronger both in aqueous environments and at ambient conditions (300 K); in water environments at neutral pH, trivalent and pentavalent methylarsenicals will be adsorbed mainly in their neutral and anionic forms, respectively. Therefore, SiG emerges as a promising material for technologies related to the removal of organic arsenic pollutants in their trivalent and pentavalent oxidation states.

---

1. Introduction

Nowadays, arsenic compounds correspond to a priority class of pollutants whose presence in natural sources constitutes a critical environmental problem; the WHO (World Health Organization) established the standards for arsenic in drinking water and air quality as 10 μg L⁻¹ and 1.5 × 10⁻³ per μg m⁻³, respectively.^1,2 Nevertheless, these standards are exceeded in many places of the planet due to the anthropogenic and industrial activities, or by the geochemical nature of the ground. The methylated forms of arsenic appear as part of the arsenic cycle in the environment, or from the arsenic metabolism in humans and animals.³ The exposure to methylated arsenic containing waters has been associated with alterations in the birth weight and gestational age,⁴ skin lesions,⁵ lung and bladder cancer,⁶ renal dysfunctions,⁷ among other effects on the human health. Due to the high cellular uptake of the trivalent methylarsenicals, their toxicity is even larger than those determined for inorganic trivalent arsenic (iAs^III).^8,9 Moreover, although the cellular uptake is lower for the pentavalent methylarsenicals with respect to inorganic pentavalent arsenic (iAs^V), their toxicity is dangerous in long periods, and they are also reported as carcinogens.^10,11 Therefore, the development of new materials for the control and removal of methylarsenicals from the environment becomes an important effort.

In recent years, graphene has emerged as a useful material for the development of new adsorbent materials for extraction of several organic and inorganic pollutants;^12,13 due to that its lamellar structure, graphene allows a large surface area available for pollutant adsorption, improving the performance, for instance, in wastewater treatment.^12–14 For the arsenic removal, intrinsic graphene has been determined with low efficiency from water sources because of the weak adsorbent–adsorbate interactions;^13,15 however, modified graphene (and oxidized graphene) composites have been reported as efficient adsorbent materials for the removal and/or detection of arsenic compounds, with high adsorption capacities from drinking waters and low adsorbent aggregation.^16–21 In this regard, density functional theory (DFT) studies have shown the ability of silicon embedded graphene (SiG) for the adsorption of a series of molecules such as hydrogen cyanide,²² nitrogen oxides,^23,24 formic acid,²⁵ oxygen,^24,26 water,²⁴ and carbon monoxide;^24,27 besides, the silicon doping increases the hydrogen storage in graphene.²⁸ Furthermore, we have recently shown from DFT calculations that SiG also shows good adsorbent properties for the removal of inorganic trivalent arsenic even in water environments,¹⁵ with adsorption energies of ∼1.0 eV.

Although only a few experimental efforts were developed to obtain silicon doped graphene, some recent works have reported its successful obtaining. Lv and co-workers achieved the chemical synthesis of large-area SiG sheets from methoxytrimethylsilane (silicon precursor) and hexane (carbon precursor), which were grown over copper substrates; the SiG sheets showed an enhanced performance for the adsorption and sensing and the Raman scattering of organic dyes.²⁹ In addition, Wang and co-workers obtained SiG by means of the chemical vapour deposition technique, using triphenylsilane as a carbon and silicon precursor; SiG was also grown onto a copper substrate.³⁰ In this sense, new technological applications of SiG are expected to be announced in the coming years.

Taking into account the rise of SiG and its potential applications, and the development of new adsorbent materials for arsenic removal, we study by means of DFT calculations the removal of methylated arsenic pollutants using SiG sheets. The adsorption of trivalent and pentavalent forms of methylated arsenicals onto SiG was fully characterized from its geometrical parameters, energetic and binding properties. Solvent effects and molecular dynamics studies were also performed to account for the removal efficiency at water and ambient conditions.

2. Computational methodology

A C₉₄H₂₄ finite lattice was adopted as a graphene surface model, where the dangling bonds at the edges were saturated with hydrogen atoms; the silicon doped graphene (SiG) was built by replacing one carbon atom by one Si atom (SiC₉₃H₂₄). The adsorption energies were well converged onto this adsorbent model, so that it was selected for the calculations. The dopant concentration was retained below 5%, because only a silicon doping content of ∼1.75–2.63% has been experimentally reached.^29,30 As trivalent methylarsenicals, we study the water soluble trivalent species monomethylarsonous acid [MMA^III, CH₃As(OH)₂] and dimethylarsinous acid [DMA^III, (CH₃)₂AsOH], which are more toxic than the inorganic form, this is arsenous acid [iAs^III, As(OH)₃]. As pentavalent methylarsenicals, we use monomethylarsonic acid [MMA^V, CH₃AsO(OH)₂] and dimethylarsinic acid [DMA^V, (CH₃)₂AsO(OH)], and comparisons were performed with inorganic pentavalent arsenic or arsenic acid [iAs^V, AsO(OH)₃]. Neutral and anionic forms of pentavalent arsenicals were studied, where the anionic forms of iAs^V, MMA^V and DMA^V are called arsenate, methylarsenate and dimethylarsenate, respectively.

All the DFT calculations were performed in the ORCA 3.0 program³¹ with the Perdew–Burke–Ernzerhof (PBE) functional;³² the usage of the PBE method has been reported to be adequate in the study of the arsenic adsorption onto mineral substrates such as iron and titanium oxides,^33,34 and even onto intrinsic and doped graphene in good agreement with more expensive ab initio methodologies.¹⁵ The all-electron Ahlrichs polarized double zeta basis sets (SVP) were used for H, C, O and Si atoms; a polarized triple-zeta basis set (TZVP) was adopted for As. The pairwise dispersion correction DFT-D3 was used to include dispersion force effects on energy and gradients, in combination with the Becke–Johnson damping scheme in order to avoid repulsive interatomic forces at short distances.^35,36 Geometry optimizations and SCF steps were converged with tolerance values of 1 × 10⁻⁶ and 1 × 10⁻⁸ Hartree, respectively. The COSMO solvent model was implemented in selected calculations,³⁷ with water as solvent; the COSMO calculations were carried-out in the ORCA 3.0 program.

Atomic net charges were obtained from the natural population analysis performed in the NBO 6.0 program;³⁸ in these cases, single point calculations were performed in the ORCA 3.0 program with the NPA keyword in order to obtain the .47 file for all the systems; the .47 file contains information about the basis set, the density matrix, the Fock matrix, among others, which are used by the NBO 6.0 program to obtain properties as natural charges and natural bond orbitals. Wavefunction and Atoms-in-Molecules (AIM) analyses were carried-out in the Multiwfn analyzer;³⁹ in these cases, the .gbw files (which are generated from the ORCA calculations) were converted to .molden files by means of the orca_2mkl program included in the ORCA 3.0 package; then, the .molden files were used as input files for the Multiwfn code.

For all the adsorbent–adsorbate systems, the adsorption energies (E_ads) were obtained as:

E_ads = E_adsorbent + E_P − E_adsorbent+P + E_vdW (1)

where, E_adsorbent, E_P and E_adsorbent+P are the total energies of the adsorbent, pollutant, and adsorbent–pollutant systems, respectively. Positive values of E_ads indicate the stability of the adsorbent–adsorbate systems. In addition, E_vdW corresponds to the contribution of dispersion forces (van der Waals interactions) to the adsorption energies. The counterpoise method⁴⁰ was used to include corrections for the basis set superposition errors (BSSEs).

The adsorbate–adsorbent stability was analyzed by means of ab initio molecular dynamic (MD) calculations with the ADMP method (atom density matrix propagation),^41–43 which allows to obtain the propagation of both the nuclear centers and electron density; thus, giving an adequate behavior of the chemical bonding in dynamic conditions. The potential was determined “on-the-fly” at the PBE/3-21G* level of theory (a minimal basis set was implemented due to the high computational cost). The time step for all the simulations was of 0.1 fs, and 1.0 ps was used for statistics; control of temperature was reached by velocity scaling. ADMP-MD calculations were performed in the Gaussian09 program⁴⁴ from optimized molecular structures in the ORCA 3.0 program.

3. Results and discussion

3.1 Adsorption of MMA^III and DMA^III

In the first place, the trivalent arsenicals are deprotonated above pH = 9.2; then, trivalent arsenicals are neutrally charged at neutral and physiological pH.^3,35,36 For that reason, the neutral forms of MMA^III and DMA^III were selected for the calculations (Fig. 1a).

Fig. 1 (a) Optimized molecular structures of MMA^III and DMA^III; distances in angstroms (Å); color code by atom: white (H); grey (C); red (O); purple (As). (b) General sketch for the interaction of trivalent methylarsenicals onto SiG; R₁ and R₂ groups change between MMA^III and DMA^III; green lines stand for double bonds, and hydrogen atoms were omitted in the molecular representation.

In Fig. 1 are depicted the optimized molecular structures of the SiG–MMA^III and SiG–DMA^III systems, where three (a–c) and two (a and b) stable conformations were obtained, respectively. Two geometrical orientations were found; (i) where the pollutant acquires a “lying-down” orientation onto SiG (conformation a), with all the As–OH and As–CH₃ bonds directed to the substrate plane; and (ii) where the pollutant acquires a “seated” orientation onto SiG (conformations b and c), with one As–OH (or As–CH₃) bond directed away from the adsorbent plane. In all the cases, the adsorbent–adsorbate interactions take place through a Si–O bond (chemisorption), with the bond lengths in the range of d_Si–O1 = 1.83–1.88 Å (Fig. 2). No interactions were obtained with the arsenic moiety, with Si⋯As distances over 3.3 Å, which agree with the substrate–As distances in the arsenic adsorption onto mineral substrates (3.2–3.5 Å) as determined from EXAFS experiments.^45,46 In addition, in the free MMA^III and DMA^III, the As–OH and As–CH₃ bonds show distances of ∼1.8 and ∼2.0 Å, respectively; although the As–CH₃ bonds remain almost unchanged after the chemisorption, the interacting As–OH bond is elongated in up to ∼0.3 Å, thus, slightly destabilizing the pollutant structure.

Fig. 2 Molecular structures of MMA^III and DMA^III adsorbed onto SiG; up to three conformations were obtained per system (a–c). Distances are in angstroms (Å). Color code: white (H); grey (C); red (O); sky-blue (Si); purple (As).

With regard to the adsorption strength, in Table 1 are listed the adsorption energies (E_ads) of the different SiG–pollutant conformations. As observed in Table 1, MMA^III and DMA^III show similar E_ads values, which are ranging from 0.9 to 1.1 eV; there are not significant differences between the adsorption strength of the different conformations, with differences of up to 0.15 eV. The adsorption strength is similar to that of inorganic trivalent arsenic (iAs^III) onto SiG (∼1.0 eV),¹⁵ indicating that the adsorption of either iAs^III, MMA^III or DMA^III will be stable onto SiG. It is worth noting that in addition to the chemical binding, the dispersion forces contribute to the stability of all these systems (E_vdW and Q_P→G parameters in Table 1). The contribution of dispersion forces is ∼0.4 eV, which represents up to 42% of the adsorption energy (∼0.4 eV). It was also observed that the chemisorption process is accompanied by an electron donation of ∼0.2|e| from the methylarsenicals towards the SiG sheet.

Table 1 Properties of MMA^III and DMA^III adsorbed onto SiG: adsorption energy (E_ads); contribution of non-covalent interactions to E_ads (E_vdW); pollutant (P) → adsorbent (G) electron transfer after the adsorption (Q_P→G); silicon–oxygen d_Si–O1 and arsenic–oxygen d_As–O1 bond lengths. Distances are in angstroms (Å); energies are in electron volt (eV)

Pollutant		Eads	EvdW	QP→G	dSi–O1	dAs–O1
a Adsorption energies determined by the B3LYP-D3 method.
MMA^III	a	1.06 (1.08)^a	0.43	0.19	1.88	2.03
	b	1.09 (1.13)	0.38	0.18	1.85	2.05
	c	0.94 (1.00)	0.40	0.15	1.83	2.10
DMA^III	a	1.07 (1.10)	0.42	0.17	1.87	1.99
	b	0.98 (1.04)	0.37	0.15	1.86	2.01
H2O		0.46	0.11		2.07

---

Another relevant point is that the reliability of the PBE adsorption energies was evaluated against those obtained with the hybrid B3LYP-D3 method in combination with the same basis sets as implemented for PBE. These results are displayed in Table 1 (in parenthesis), and it was found that the differences between both methods are less than 0.10 eV, indicating the accuracy of the selected methodology without a significant increase in the computational cost.

At this point, it is important to highlight that the adsorption of trivalent arsenicals onto the SiG graphene shows a reasonable performance in comparison with mineral substrates. For instance, Oliveira and co-workers⁴⁷ determined the arsenite adsorption onto aluminum oxides (gibbsite) with adsorption energies of ∼0.64 eV from DFT calculations; Blanchard and co-workers³³ showed from DFT calculations that the bidentate interaction of arsenite with FeS₂ takes place with an adsorption energy of ∼1.44 eV; moreover, Wei and co-workers⁴⁸ has determined the arsenite adsorption onto TiO₂ anatase (which is a well know to effectively remove arsenic from polluted waters) with adsorption energies of up to ∼0.80 eV at neutral conditions. In summary, SiG show good adsorbent properties for the removal of organic and inorganic trivalent methylarsenicals. On the other hand, the industrial activity is the main source of atmospheric arsenic and responsible for the presence of methylarsenicals in the particulate matter,^49–51 so that the proposed adsorbents could be also applied for removal of methylarsenicals in the gas phase or its detection.

3.2 Adsorption of MMA^V and DMA^V

Additionally, we also studied the ability of SiG to adsorb pentavalent methylarsenicals (MMA^V, DMA^V). Also, the adsorption of inorganic pentavalent arsenic (iAs^V) was studied for comparison purposes and completeness of the work. It is worth to note that the pentavalent arsenicals have different acid–base equilibrium depending on the pH. For instance, iAs^V is a triprotic acid and dissociates in water according to:^3,52

AsO(OH)₃ ⇔ AsO₂(OH)₂⁻ + H⁺, pK_a1 = 2.3 (2)

AsO₂(OH)₂⁻ ⇔ AsO₃(OH)²⁻ + H⁺, pK_a1 = 6.8 (3)

AsO₃(OH)²⁻ ⇔ AsO₄³⁻ + H⁺, pK_a1 = 11.6 (4)

Therefore, almost equal forms of arsenate AsO₂(OH)₂⁻ and AsO₃(OH)²⁻ could exist at neutral pH. Likewise, MMA^V and DMA^V behave as diprotic and monoprotic acids, respectively, with the equilibrium:^3,53

CH₃AsO(OH)₂ ⇔ CH₃AsO₂(OH)⁻ + H⁺, pK_a1 = 4.1 (5)

CH₃AsO₂(OH)⁻ ⇔ CH₃AsO₃²⁻ + H⁺, pK_a1 = 8.7 (6)

(CH₃)₂AsO(OH) ⇔ (CH₃)₂AsO₂⁻ + H⁺, pK_a1 = 6.2 (7)

Consequently, the anionic form of MMA^V [CH₃AsO₂(OH)⁻] is the main one at neutral pH; conversely, both neutral and anionic species of DMA^V (CH₃)₂AsO(OH) and (CH₃)₂AsO₂⁻ are capable to exist at neutral pH. Taking into account these considerations, we performed calculations onto the anionic forms, which are expected to be the main species adsorbed onto SiG at neutral conditions in aqueous environments (Fig. 3a). The neutral forms were also computed because they are important to get insights about the removal ability in a broad range of pH, as well as for the arsenic adsorption in the gas phase.

Fig. 3 (a) Optimized molecular structures of iAs^V, MMA^V and DMA^V at their anionic and neutral forms; distances in angstroms (Å); color code by atom: white (H); grey (C); red (O); purple (As). (b) General sketch for the interaction of pentavalent arsenicals onto SiG; R₁ and R₂ groups change between iAs^V, MMA^III and DMA^III; green lines stand for double bonds, and hydrogen atom were deleted to simplify.

In the first place, the optimized molecular structures of the anionic MMA^V and DMA^V adsorbed onto SiG are displayed in Fig. 4, in addition to those for iAs^V for comparison. Two (a and b), three (a–c) and two (a and b) conformations were found for the adsorption of iAs^V, MMA^V and DMA^V, respectively, where a “seated” orientation is adopted. In all the cases, silicon binds to one deprotonated oxygen atom of the pollutant (which is labeled as O₁ in Fig. 3b), with the bond lengths of d_Si–O ≈ 1.7 Å; note that these distances are at least ∼0.1 Å shorter than in the adsorption of the trivalent forms. The latter suggest that the adsorption strength could be sorted as DMA^V > MMA^V > iAs^V, and even that the adsorption energies would be higher than for trivalent methylarsenicals. Indeed, from Table 2, it is noted an important increasing in the adsorption energies, with values of up to 4.01 eV. Specifically, the chemisorption of iAs^V, MMA^V and DMA^V is characterized with E_ads values on the range of 3.31–3.34 eV, 3.63–3.64 eV and 3.93–4.01 eV, respectively, which agree with the slight differences in the Si–O₁ bond lengths, and sorting the chemisorption strength as DMA^V > MMA^V > iAs^V (in other words, the adsorption strength increases as the methyl groups in the pollutant increases). Note that the interacting As–O₁ bond distances in the anionic MMA^V and DMA^V are slightly affected due to the formation of the Si–O₁ interaction, with the bond lengths of d_As–O1 ≈ 1.80 Å; the latter indicates that the σ_As–O1 bond is not weakened due to the chemisorption as observed in the case of MMA^III/DMA^III, and retaining the pollutant stability.

Fig. 4 Molecular structures of iAs^V, MMA^V, DMA^V and DMMTA^V (in their anionic forms) adsorbed onto SiG; up to four conformations were obtained per system (a–d). Distances are in angstroms (Å). Color code: white (H); grey (C); red (O); sky-blue (Si); yellow (S); purple (As).

Table 2 Properties of pentavalent arsenicals adsorbed onto SiG: adsorption energy (E_ads); contribution of dispersion forces to E_ads (E_vdW); pollutant (P) → adsorbent (G) electron transfer after the adsorption (Q_P→G); silicon–oxygen d_Si–O1 and arsenic–oxygen d_As–O1 bond lengths. Distances are in angstroms (Å); energies are in electron volt (eV)

Pollutant		Eads	EvdW	QP→G	dSi–O1	dAs–O1
a Adsorption energies computed with the minimally augmented def2 basis sets (ma-def2-SVP).b For DMMTA^V, these interaction modes correspond to Si–S binding.c In these conformations of DMMTA^V, this bond corresponds to As–S.d In these cases, the interaction with silicon is with the As=O2 moiety as showed in Fig. 3b.
Anionic pentavalent arsenicals
iAs^V	a	3.31	0.35	0.40	1.73	1.78
	b	3.33	0.37	0.41	1.72	1.77
MMA^V	a	3.64 (3.76)^a	0.40	0.42	1.71	1.79
	b	3.64 (3.70)^a	0.36	0.41	1.72	1.78
	c	3.63 (3.67)^a	0.36	0.42	1.72	1.78
DMA^V	a	4.01 (4.14)^a	0.42	0.43	1.70	1.79
	b	3.93 (4.04)^a	0.38	0.42	1.71	1.80
DMMTA^V	a	3.79	0.43	0.42	1.70	1.80
	b	3.72	0.43	0.41	1.71	1.80
	c	2.96	0.47	0.68	2.20^b	2.28^c
	d	3.08	0.49	0.72	2.19^b	2.28^c
Neutral pentavalent arsenicals
iAs^V	a	0.67	0.44	0.16	1.89	1.95
	b	1.52	0.45	0.19	1.73^d	1.82^d
MMA^V	a	0.71	0.44	0.20	1.89	1.97
	b	0.58	0.40	0.19	1.94	1.95
	c	0.57	0.43	0.16	1.89	1.96
	d	1.58	0.47	0.23	1.79^d	1.73^d
	e	1.66	0.44	0.24	1.79^d	1.73^d
DMA^V	a	1.23	0.45	0.20	1.88	1.97
	b	1.73	0.48	0.23	1.78^d	1.74^d
	c	1.77	0.45	0.26	1.79^d	1.73^d

---

With respect to the contribution of dispersion forces (E_vdW) to the adsorption strength, it can be observed in Table 2 that they are not so important, with contributions of up to 0.49 eV (∼10% of the total E_ads). On the other hand, it is clear that the anionic nature of MMA^V and DMA^V increases the amount of electron transfer towards the acceptor SiG in up to 100% with respect to the trivalent forms. The electron transfer is similar for each the arsenic compounds, where ∼0.4|e| are transferred to the adsorbent.

It must be also noted that due to the anionic nature of the pollutants, the accuracy of the adsorption energies was evaluated against those obtained by using a diffuse basis set. The minimally augmented Karlsruhe def2 basis sets (ma-def2-SVP)⁵⁴ were used, which include diffuse s and p functions for all the atoms; these calculations were performed only with anionic MMA^V and DMA^V due to the high computational cost. Note that other diffuse basis sets showed problems with linear dependences and convergence. From Table 2 (in parenthesis), we observe that only differences of up to 0.13 eV are obtained in the E_ads values comparing both basis sets, indicating the good accuracy of the E_ads values.

Dimethylmonothioarsinic acid (DMMTA^V) is another common methylated pollutant, whose presence in the environment is favored by the metabolic process of mammals, in addition to the sulfidic nature of grounds.^55–58 The cytotoxicity of DMMTA^V is larger than DMA^V, and similar to DMA^III, and its toxicity has been associated with the production of reactive oxygen molecules during its exposure, which could cause DNA damage in order to initialize cancer.^3,10 Taken into account these considerations, we carried-out calculations with anionic DMMTA^V (dimethylmonothioarsenate) as adsorbate to get insights into the removal of methylated thioarsenicals. The SiG–DMMTA^V conformations are also displayed in Fig. 4, and their properties are quoted in Table 2. Four interaction modes were found (a–d), where the chemisorption takes place by means of Si–O (a and b) or Si–S (c and d) binding, with bond distances of d_Si–O ≈ 1.7 Å and d_Si–S ≈ 2.2 Å, respectively. The chemisorption is favored in the Si–O mode with E_ads values of up to 3.79 eV; while, the Si–S conformations reach E_ads value of up to 3.08 eV. Consequently, the efficiency of SiG would be good for a broad kind of trivalent and pentavalent arsenicals, also including their thiolated forms.

We studied the adsorption of neutral iAs^V, MMA^V and DMA^V to get insights about the removal at low pH or gas phase (see Fig. 5 and Table 2). In these cases, the chemisorption takes place by two binding modes. In the first binding mode (conformations a, a–c, and a for iAs^V, MMA^V and DMA^V, respectively), the silicon atom interacts with the protonated oxygen belonging to one σ_As–O1 bond of the pollutant, with bond distances of d_Si–O1 ≈ 1.90 Å. The adsorption energies reach values of up to 0.67, 0.71 and 1.23 for iAs^V, MMA^V and DMA^V, respectively, which are similar in comparison with adsorption of trivalent methylarsenicals. Note that in these conformations, the As–O₁ bond it is elongated from −1.8 to −2.0 Å, suggesting the depletion of the σ_As–O1 bond as for trivalent arsenicals.

Fig. 5 Molecular structures of iAs^V, MMA^V and DMA^V (in their neutral forms) adsorbed onto SiG; up to five conformations were obtained per system (a–e). Distances are in angstroms (Å). Color code: white (H); grey (C); red (O); sky-blue (Si); purple (As).

Conversely, in the second binding mode, silicon binds to the deprotonated oxygen atom belonging to the π_As=O2 bond of the pollutant (conformations c–e, and b and c for iAs^V, MMA^V and DMA^V, respectively). These conformations are favored with short bond lengths in the range of d_Si–O1 = 1.73–1.79 Å, in addition to enhanced adsorption energies of up to 1.52, 1.66 and 1.77 eV for iAs^V, MMA^V and DMA^V, respectively. In these cases, the double π_As=O2 bond is elongated from ∼1.63 to ∼1.73 Å for MMA^V and DMA^V, which is near to distances for a single As–O bond. Then, the π_As=O2 bond would be weakened to allow the Si–O₂ interaction, but without a major structural destabilization of the pollutant because of the As–O₂ bond remains still stable by single bonding. Likewise, in the case of iAs^V, the presence of three electron-withdrawing –OH groups in the pollutant structure cause a slight larger depletion of the π_As=O2 bond density with respect to the methylated arsenicals, increasing its bond distance from 1.62 to 1.82 Å. Finally, these results indicate that the adsorption strength of neutral arsenicals onto SiG can be sorted as DMA^V > MMA^V > iAs^V > DMA^III ≈ MMA^III ≈ iAs^III; this finding might to serve as a guidance for experimental studies.

In summary, the adsorption of pentavalent methylarsenicals (inorganic and organic) appears highly improved onto SiG with respect to the trivalent forms, so that a promising new material for arsenic removal is proposed with efficiency from neutral to low pH conditions. The latter is also in agreement with the lower mobility of pentavalent arsenicals in comparison with those obtained for the trivalent species when they are adsorbed from aqueous solutions, which also causes the high toxicity of the trivalent forms.^59,60 As noted above, the anionic and neutral pentavalent methylarsenicals (also including arsenate) are adsorbed with energies of up to 4.01 and 1.77 eV, respectively, which constitutes an important improvement in comparison to other adsorbents. For instance, the arsenate adsorption onto TiO₂ based surfaces (such as anatase and rutile) has been determined with adsorption energies of up to ∼3.0 eV;^34,61,62 Ladeira and co-workers estimated in ∼2.6 eV the arsenate chemisorption onto gibbsite (an aluminum hydroxide);⁴⁵ the arsenate adsorption onto mineral iron oxides (such as goethite, lepidocrocite and hematite) reached a value of ∼2.4 eV;⁶³ while, iAs^V and DMA^V are chemisorbed onto Fe-(oxyhydr)oxides with adsorption energies of up to ∼1.6 eV.⁶⁴

With regard to the recovery of SiG after the removal, it is known that the adsorption of anionic pollutants is competitive in alkaline conditions with the adsorption of hydroxyl anions; the latter avoids the uptake of anionic pollutants because the hydroxyl groups turn negatively charged the adsorbent surface, and causing a repulsive coulombic interaction with the anionic arsenicals.¹³ Then, the recovery of the adsorbent material would be straightforwardly reached by means of high pH solutions. On the other hand, a high recovery of doped graphene adsorbents has been reached by thermal treatment; for instance, thermal treatment up to 800 °C allows the desorption of adsorbed thiophene (84%) from P-doped graphene, without thermal degradation of the substrate.⁶⁵

3.3 Atoms in molecules (AIM) analysis

In this section, we present some interesting features determining the strength of the SiG–pollutant interactions. We used the electron density (ρ_i) at the bond critical points (BCPs) as obtained from the AIM methodology, which are listed in Table 3. To take into consideration, note that due to the different electronegativity values of C, O and As atoms [2.55 (C), 3.44 (O) and 2.18 (As) in the Pauling scale], the electron bond densities at the BCP of the As–C and As–O bonds (ρ_Si–O1 and ρ_As–C) are lower than 0.2 a.u. because the bonds have a relative high ionic character;⁶⁶ in addition, the Si–O bond corresponds to a closed-shell interaction,⁶⁶ so that the electron density at the BCP will be of ∼0.1 a.u.

Table 3 Electron density at the BCPs (ρ_i) of selected bonds in the SiG–pollutant systems; ρ_i values in a.u.; ρ_C⋯H and ρ_C⋯H′ correspond to weak BCPs assigned to C⋯H interactions

Pollutant		ρSi–O1	ρAs–O1	ρC⋯H	ρC⋯H′
Neutral trivalent arsenicals
MMA^III	a	0.074	0.086	0.010	0.023
	b	0.079	0.082	0.023	—
	c	0.082	0.077	0.013	0.020
DMA^III	a	0.074	0.091	0.008	0.013
	b	0.077	0.087	0.012	—
Anionic pentavalent arsenicals
MMA^V	a	0.107	0.145	0.020	0.011
	b	0.106	0.148	0.013	—
	c	0.105	0.149	0.005	0.007
DMA^V	a	0.109	0.142	0.008	0.011
	b	0.107	0.143	0.007	0.005
Neutral pentavalent arsenicals
MMA^V	a	0.071	0.098	0.028	0.008
	b	0.066	0.104	0.021	—
	c	0.070	0.102	0.008	—
	d	0.089	0.169	0.029	0.008
	e	0.088	0.171	0.026	0.019
DMA^V	a	0.072	0.097	0.015	0.006
	b	0.091	0.164	0.008	0.008
	c	0.088	0.169	0.023	0.013

---

First, when MMA^III and DMA^III are adsorbed onto SiG, the interacting As–O₁H bond is elongated up to ∼0.3 Å. In this regard, the electron density at the BCP of the As–O₁ bond (ρ_As–O1) shows values of 0.08–0.09 a.u., indicating that the single bond is weakened in up to ∼40% due to the chemisorption with respect to the free pollutants (ρ_As–O1 ≈ 0.14 a.u.). In addition, the electron density at the BCP of the Si–O₁ bond are on the range of ρ_Si–O1 = 0.07–0.08 a.u., which are lower than for a single covalent bond due to the coordinative covalent character of this interaction. These results clearly suggest that the in the SiG–pollutant interaction, the pollutant mainly transfers electron density from the interacting As–O₁ bond toward the new formed Si–O₁ bond, causing the destabilization of the pollutant structure and limiting the adsorption strength.

With regard to the free MMA^V and DMA^V, the deprotonation of one –OH group causes the improving and weakening of the σ_As–O1 and π_As=O2 bonds, respectively, due to resonance effects. For instance, from Fig. 3a, it is observed that for the neutral DMA^V molecule, the As–O₁ and As–O₂ bond distances are about 1.80 and 1.64 Å, respectively, showing a double and single bond, respectively; whereas, in the free anionic DMA^V, the As–O₁ and As–O₂ bonds are equivalents with distances of 1.67 Å and bond order of ∼1.73, so that the electron density excess is delocalized trough the σ_As–O1 and π_As=O2 bonds to form two resonant double bonds.

The above has an impact on the adsorption strength of anionic pentavalent arsenicals onto SiG (and probably onto other substrates). In these cases, the ρ_As–O1 values after the chemisorption remain in ∼0.14 a.u., indicating that the bond turns more covalent than ionic and retains the single bond nature unlike in the adsorption of trivalent arsenicals. The negative charge excesses also increase the stability of the Si–O₁ bond, increasing its density up to ρ_Si–O1 ≈ 0.11 a.u. Therefore, the high adsorption energies obtained for the chemisorption of anionic MMA^V and DMA^V are explained because the interacting As–O₁ bond of the pollutant is strengthened before the interaction due to resonant effect, showing a quasi double bonding; then, when chemisorption is reached, the quasi double As–O₁ bond transfer bond density to strengthen the Si–O₁ bonding, but the final As–O₁ bond is capable to retain a single nature.

Finally, in the case of neutral pentavalent arsenicals, the most stable conformations take place by binding of silicon with the deprotonated oxygen atom belonging to the pollutant π_As=O2 bond (conformations d and e for MMA^V, and b and c for DMA^V). In these cases, the electron density at the BCP indicates that the double π_As=O2 bond is weakened to allow the Si–O₂ interaction without a major structural destabilization of the pollutant because of the As–O₂ bond remains still stable by single bonding, thus causing a higher adsorption energy with respect to the interaction by the σ_As–O1 bond (in these cases the behavior is similar to trivalent methylarsenicals). For instance, in conformation “c” of the adsorption of neutral DMA^V onto SiG, the interacting π_As=O2 bond is weakened from the double bond in the free pollutant (ρ_As–O1 = 0.21 a.u.) until an almost single σ_As–O2 bond (ρ_As–O1 = 0.17 a.u.) after chemisorption, which helps to retain the pollutant stability, in addition to slightly increase the strength of the Si–O bond (ρ_As–O1 = 0.09 a.u.) as in the case of the anionic pentavalent methylarsenicals.

In summary, after the SiG–pollutant interactions, the stability of the σ_As–O or π_As=O bonds of the methylarsenicals plays an important role in the adsorption strength, and also explains the differences in the adsorption strength of arsenic compounds under different oxidation and charge states. Moreover, we found a clear linear relation between the adsorption energies (E_ads) and the electron bond density at the bond critical point of the Si–O bond (ρ_Si–O1) (Fig. 6), where the adsorption strength of the different methylarsenicals can be sorted as: (anionic pentavalent methylarsenicals) ≫ (neutral pentavalent methylarsenicals) > (neutral trivalent methylarsenicals), expecting that this relation could be extrapolated for a wide class of arsenic based pollutants adsorbed onto SiG.

Fig. 6 Linear relationship between the adsorption energies (E_ads) and electron density at the BCP of the Si–O bond (ρ_Si–O) for the adsorption of methylarsenicals onto SiG.

Additionally, it is noteworthy to mention that weak bond critical points were found between hydrogen atoms (belonging to the –OH or –CH₃ groups of the pollutants) and carbon atoms near to the dopant in the SiG surface; these BCPs were labeled as ρ_C⋯H and ρ_C⋯H′ in Table 3, and their values remains under ∼0.02 a.u. as expected for weak closed-shell interactions.⁶⁶ The graphs of the bond paths connecting the nuclear and bond critical points clearly indicate a weak C⋯H interaction that would to have a charge-controlled nature (Fig. 7a). The latter was confirmed from the plotting of the electron density difference (Fig. 7b), where is depicted the electron density distribution after the pollutant chemisorption. We can observe the attraction between the carbon atoms in SiG that are gaining electron density (sky-blue) and the pollutant hydrogen atoms losing electron density (blue). Therefore, the pollutant adsorption activates some carbon atoms of SiG, and these atoms electrostatically interact with the electron deficient hydrogen atoms of the pollutant, increasing the adsorbent–adsorbate stability.

Fig. 7 (a) Bond path connecting nuclear and bond critical points; the paths connecting the carbon (SiG) and hydrogen (pollutant) atoms suggest an electrostatic interaction between these atoms (see arrows). (b) Electron density difference; the electron density decreasing (yellow color) and electron density increasing (sky-blue color) sites are displayed after the adsorbent–adsorbate interaction. MMA^III and MMA^V adsorbed in “a” conformations were taken as representatives.

3.4 Solvent affects

To account for an efficient arsenic removal in water environments, it is necessary to determine if the mobility of the arsenic compounds remains low by including solvation effects; in this regard, the removal efficiency of trivalent arsenicals in aqueous environments is lower than for the pentavalent forms.⁵⁹ As noted in Table 1, the H₂O molecules are chemisorbed onto SiG with an adsorption energy of 0.46 eV per molecule, with a bond length of d_Si–O = 2.07 Å, which is stronger that the adsorption of H₂O molecules onto intrinsic graphene (<0.10 eV per molecule). Despite the latter, since the pollutant chemisorption is stronger than for H₂O molecules, SiG would serve as a superior adsorbent material for the removal of inorganic and methylated arsenic compounds. To give a theoretical basis to our last point, and to gain insights into the structural effects of water molecules onto the SiG–pollutant systems, explicit/implicit solvent calculations were done to by surrounding the adsorbent–adsorbate systems with a cluster of 15H₂O molecules in presence of the COSMO solvent model to create the “solvent environment”; then, the system where fully optimized in the ORCA 3.0 program.

As observed in Fig. 8, the water molecules surround the adsorbed pollutants, without to cause the desorption of the compounds. Additionally, the geometrical parameters of the adsorbent–adsorbate conformations are slightly affected with respect to those determined in the gas phase (see Table S1 of the ESI†); all the Si–O bond lengths appear on the range of 1.69–1.84 Å, indicating the high stability of the chemisorption. As expected for trivalent methylarsenicals (MMA^III and DMA^III), the As–O bonds are found on the range of 2.12–2.37 Å, which agrees with the destabilization of the σ_As–O bonding. Conversely, in the pentavalent methylarsenicals (MMA^V, DMA^V and DMMTA^V), the As–O bond distances appear in the range of 1.74–1.85 Å, indicating that these compounds must show a lower mobility from the adsorbent surface compared with the trivalent forms; similar results were also obtained for iAs^V. In the case of DMMTA^V, the interaction by the silicon–sulfur binding is also stable in water.

Fig. 8 Trivalent and pentavalent methylarsenicals adsorbed onto SiG in an explicit water environment (15H₂O); conformations “a” were taken as representatives. H₂O molecules are depicted in white with pointed hydrogen bonds. All the systems are included in the ESI.†

Additionally, it was obtained that the trivalent methylarsenicals remain protonated when they are adsorbed onto SiG, where the bond length of the interacting O–H group remains lower than ∼1.04 Å. On the other hand, in the pentavalent arsenicals, the H atom turns labile from the interacting O–H group, with the bond lengths above 1.79 Å, and forming H₃O⁺ molecules; the latter indicates that pentavalent arsenicals are mainly adsorbed onto SiG in their anionic forms at neutral pH conditions.

3.5 Adsorption stability

In order to account for the stability of the adsorbent–adsorbate interactions, ab initio molecular dynamics trajectories were obtained at 300 K; due to the high computational cost, these calculations were performed only for the most stable systems. We studied the stability of the Si–O₁ and As–O₁ bonds by means of the radial pair distribution of distances [g_ab(r)] (Fig. 9), which allows to obtain the distribution of bond lengths along the overall trajectory; 10 000 conformations per system were used for statistics.

Fig. 9 Radial pair distribution [g_ab(r)] of the Si–O₁ and As–O₁ distances in selected SiG–pollutant systems. 10 000 conformations per system were used for statistics, during a time t = 1.0 ps and a temperature of 300 K.

As noted in Fig. 9, the chemisorption still remains strong even in dynamic conditions, because of the Si–O₁ and As–O₁ bonds lengths are comparable to values obtained in the ground state. In the case of trivalent methylarsenicals (MMA^III and DMA^III), the Si–O₁ and As–O₁ bonds lengths are mainly retained in the range of d_Si–O1 = 1.75–1.95 Å and d_As–O1 = 1.90–2.2 Å; note that although the Si–O bond is destabilized in these cases as noted above, the bond is able to be retained. Likewise, for pentavalent methylarsenicals (MMA^V and DMA^V), the neutral and anionic forms were analyzed; in both cases the Si–O interaction is strong and ranging from d_Si–O1 = 1.69 to 1.80 Å, in agreement with the higher adsorption energies compared to the trivalent cases. However, the main differences are observed in the propagation of the As–O bonds, where the As–O bond length ranges from d_As–O1 = 1.70 to 1.90 Å for the anionic forms; while, the As–O bonds length is mainly retained in the range of d_As–O1 = 1.85–2.20 Å for the neutral forms. The latter agrees with the larger strength of the interaction of anionic arsenicals compared to its neutral forms, where differences are only given by the stability of the interacting As–O bond.

4. Conclusions

We have studied the applicability of SiG for the removal of methylarsenicals in their pentavalent and trivalent oxidation states; also comparisons with inorganic arsenic were performed. In all the cases, chemisorption is reached with high adsorption strength. In this regard, MMA^III and DMA^III are adsorbed onto SiG with adsorption energies ranging from 0.9 to 1.1 eV at neutral conditions. In the case of anionic pentavalent arsenicals (including MMA^V, DMA^V, DMMTA^V and iAs^V), the chemisorption was reached with adsorption energies in the range of 3.3–4.0 eV; while, the adsorption of their neutral forms was reached with values of up to 1.7 eV. Therefore, the adsorption strength of the different methylarsenicals adsorbed onto SiG is sorted as: (anionic pentavalent methylarsenicals) ≫ (neutral pentavalent methylarsenicals) > (neutral trivalent methylarsenicals). In addition, the adsorption strength of neutral arsenicals onto SiG was sorted as DMA^V > MMA^V > iAs^V > DMA^III ≈ MMA^III ≈ iAs^III. We also found that the stability of the σ_As–O or π_As=O bonds of the methylarsenicals plays an important role in the adsorption strength, and also explains the differences in the adsorption strength of arsenic compounds under different oxidation and charge states. Additionally, from molecular dynamics and explicit solvent simulations, it was determined that the adsorption stability remains stronger both in aqueous environments and ambient conditions at 300 K. In these cases, we determined that trivalent and pentavalent methylarsenicals will be adsorbed mainly in their neutral and anionic forms at neutral pH, respectively.

Therefore, SiG is proposed as an emergent material for technologies related to the removal of organic (and inorganic) arsenic compounds in their trivalent and pentavalent oxidation states, with an enhanced performance for removal in the gas phase, and in a broad range of pH for the removal from water sources.

Acknowledgements

This work was supported by the projects FONDECYT/Postdoctorado no. 3140314, FONDECYT no. 1130072 and ICM grant no.120082.

References

1. W. H. Organization, Guidelines for drinking-water quality: recommendations, World Health Organization, 2004.
2. W. H. O. R. O. f. Europe and W. H. Organization, Air quality guidelines: global update 2005: particulate matter, ozone, nitrogen dioxide, and sulfur dioxide, World Health Organization, 2006.
3. V. K. Sharma and M. Sohn, Environ. Int., 2009, 35, 743–759.
4. J. E. Laine, K. A. Bailey, M. Rubio-Andrade, A. F. Olshan, L. Smeester, Z. Drobná, A. H. Herring, M. Stýblo, G. G. García-Vargas and R. C. Fry, Environ. Health Perspect., 2015, 123, 186.
5. L. Yang, Y. Chai, J. Yu, B. Wei, Y. Xia, K. Wu, J. Gao, Z. Guo and N. Cui, Environ. Toxicol., 2015 DOI:10.1002/tox.22209.
6. D. Melak, C. Ferreccio, D. Kalman, R. Parra, J. Acevedo, L. Pérez, S. Cortés, A. H. Smith, Y. Yuan, J. Liaw and C. Steinmaus, Toxicol. Appl. Pharmacol., 2014, 274, 225–231.
7. B. A. Peters, M. N. Hall, X. Liu, V. Slavkovich, V. Ilievski, S. Alam, A. B. Siddique, T. Islam, J. H. Graziano and M. V. Gamble, Environ. Res., 2015, 143, 123–130.
8. E. Dopp, L. Hartmann, U. Von Recklinghausen, A. Florea, S. Rabieh, U. Zimmermann, B. Shokouhi, S. Yadav, A. Hirner and A. Rettenmeier, Toxicol. Sci., 2005, 87, 46–56.
9. E. Dopp, L. Hartmann, A.-M. Florea, U. Von Recklinghausen, R. Pieper, B. Shokouhi, A. Rettenmeier, A. Hirner and G. Obe, Toxicol. Appl. Pharmacol., 2004, 201, 156–165.
10. K. T. Kitchin, Toxicol. Appl. Pharmacol., 2001, 172, 249–261.
11. E. M. Kenyon and M. F. Hughes, Toxicology, 2001, 160, 227–236.
12. S. Wang, H. Sun, H.-M. Ang and M. Tadé, Chem. Eng. J., 2013, 226, 336–347.
13. Y. Cao and X. Li, Adsorption, 2014, 20, 713–727.
14. G. Z. Kyzas, E. A. Deliyanni and K. A. Matis, J. Chem. Technol. Biotechnol., 2014, 89, 196–205.
15. D. Cortes-Arriagada and A. Toro-Labbe, Phys. Chem. Chem. Phys., 2015, 17, 12056–12064.
16. X.-L. Wu, L. Wang, C.-L. Chen, A.-W. Xu and X.-K. Wang, J. Mater. Chem., 2011, 21, 17353–17359.
17. K. Zhang, V. Dwivedi, C. Chi and J. Wu, J. Hazard. Mater., 2010, 182, 162–168.
18. J. Zhu, R. Sadu, S. Wei, D. H. Chen, N. Haldolaarachchige, Z. Luo, J. Gomes, D. P. Young and Z. Guo, ECS J. Solid State Sci. Technol., 2012, 1, M1–M5.
19. D. Nandi, K. Gupta, A. K. Ghosh, A. De, S. Banerjee and U. C. Ghosh, in Nanotechnology for Sustainable Development, Springer, 2014, pp. 149–162.
20. G. Sheng, Y. Li, X. Yang, X. Ren, S. Yang, J. Hu and X. Wang, RSC Adv., 2012, 2, 12400–12407.
21. T. Wen, X. Wu, X. Tan, X. Wang and A. Xu, ACS Appl. Mater. Interfaces, 2013, 5, 3304–3311.
22. S. F. Rastegar, A. A. Peyghan and N. L. Hadipour, Appl. Surf. Sci., 2013, 265, 412–417.
23. Y. Chen, B. Gao, J.-X. Zhao, Q.-H. Cai and H.-G. Fu, J. Mol. Model., 2012, 18, 2043–2054.
24. Y. Zou, F. Li, Z. Zhu, M. Zhao, X. Xu and X. Su, Eur. Phys. J. B, 2011, 81, 475–479.
25. M. D. Esrafili, R. Nurazar and E. Vessally, Int. J. Quantum Chem., 2015, 115, 1153–1160.
26. Y. Chen, X.-c. Yang, Y.-j. Liu, J.-x. Zhao, Q.-h. Cai and X.-z. Wang, J. Mol. Graphics Modell., 2013, 39, 126–132.
27. R. Gholizadeh and Y.-X. Yu, Appl. Surf. Sci., 2015, 357, 1187–1195.
28. J. H. Cho, S. J. Yang, K. Lee and C. R. Park, Int. J. Hydrogen Energy, 2011, 36, 12286–12295.
29. R. Lv, M. C. dos Santos, C. Antonelli, S. Feng, K. Fujisawa, A. Berkdemir, R. Cruz-Silva, A. L. Elías, N. Perea-Lopez and F. López-Urías, Adv. Mater., 2014, 26, 7593–7599.
30. Z. Wang, P. Li, Y. Chen, J. Liu, W. Zhang, Z. Guo, M. Dong and Y. Li, J. Mater. Chem. C, 2015, 3, 6301–6306.
31. F. Neese, Wiley Interdiscip. Rev.: Comput. Mol. Sci., 2012, 2, 73–78.
32. J. P. Perdew, K. Burke and Y. Wang, Phys. Rev. B: Condens. Matter Mater. Phys., 1996, 54, 16533.
33. M. Blanchard, K. Wright, J. D. Gale and C. R. A. Catlow, J. Phys. Chem. C, 2007, 111, 11390–11396.
34. Z. Wei, K. Liang, Y. Wu, Y. Zou, J. Zuo, D. C. Arriagada, Z. Pan and G. Hu, J. Colloid Interface Sci., 2016, 462, 252–259.
35. S. Grimme, J. Antony, S. Ehrlich and H. Krieg, J. Chem. Phys., 2010, 132, 154104.
36. S. Grimme, S. Ehrlich and L. Goerigk, J. Comput. Chem., 2011, 32, 1456–1465.
37. A. Klamt, Wiley Interdiscip. Rev.: Comput. Mol. Sci., 2011, 1, 699–709.
38. E. D. Glendening, C. R. Landis and F. Weinhold, J. Comput. Chem., 2013, 34, 1429–1437.
39. T. Lu and F. Chen, J. Comput. Chem., 2012, 33, 580–592.
40. S. F. Boys and F. Bernardi, Mol. Phys., 1970, 19, 553–566.
41. H. B. Schlegel, J. M. Millam, S. S. Iyengar, G. A. Voth, A. D. Daniels, G. E. Scuseria and M. J. Frisch, J. Chem. Phys., 2001, 114, 9758–9763.
42. S. S. Iyengar, H. B. Schlegel, J. M. Millam, G. A. Voth, G. E. Scuseria and M. J. Frisch, J. Chem. Phys., 2001, 115, 10291–10302.
43. H. B. Schlegel, S. S. Iyengar, X. Li, J. M. Millam, G. A. Voth, G. E. Scuseria and M. J. Frisch, J. Chem. Phys., 2002, 117, 8694–8704.
44. M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, G. Scalmani, V. Barone, B. Mennucci, G. A. Petersson, H. Nakatsuji, M. Caricato, X. Li, H. P. Hratchian, A. F. Izmaylov, J. Bloino, G. Zheng, J. L. Sonnenberg, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, T. Vreven, J. A. Montgomery, J. E. Peralta, F. Ogliaro, M. Bearpark, J. J. Heyd, E. Brothers, K. N. Kudin, V. N. Staroverov, R. Kobayashi, J. Normand, K. Raghavachari, A. Rendell, J. C. Burant, S. S. Iyengar, J. Tomasi, M. Cossi, N. Rega, J. M. Millam, M. Klene, J. E. Knox, J. B. Cross, V. Bakken, C. Adamo, J. Jaramillo, R. Gomperts, R. E. Stratmann, O. Yazyev, A. J. Austin, R. Cammi, C. Pomelli, J. W. Ochterski, R. L. Martin, K. Morokuma, V. G. Zakrzewski, G. A. Voth, P. Salvador, J. J. Dannenberg, S. Dapprich, A. D. Daniels, Ö. Farkas, J. B. Foresman, J. V. Ortiz, J. Cioslowski, D. J. Fox, Rev B01, Gaussian Inc., Wallingford CT, 2010.
45. A. Ladeira, V. Ciminelli, H. Duarte, M. Alves and A. Ramos, Geochim. Cosmochim. Acta, 2001, 65, 1211–1217.
46. G. Morin, G. Ona-Nguema, Y. Wang, N. Menguy, F. Juillot, O. Proux, F. Guyot, G. Calas and G. E. Brown Jr, Environ. Sci. Technol., 2008, 42, 2361–2366.
47. A. F. Oliveira, A. C. Q. Ladeira, V. S. Ciminelli, T. Heine and H. A. Duarte, J. Mol. Struct.: THEOCHEM, 2006, 762, 17–23.
48. Z. Wei, S. Zhang, Z. Pan and Y. Liu, Appl. Surf. Sci., 2011, 258, 1192–1198.
49. H. Mukai and Y. Ambe, Atmos. Environ., 1987, 21(1967), 185–189.
50. T. Tziaras, S. A. Pergantis and E. G. Stephanou, Environ. Sci. Technol., 2015, 49, 11640–11648.
51. D. Sánchez-Rodas, A. M. S. de la Campa, D. Jesús, V. Oliveira, J. L. Gómez-Ariza, X. Querol and A. Alastuey, Chemosphere, 2007, 66, 1485–1493.
52. S. Goldberg and C. T. Johnston, J. Colloid Interface Sci., 2001, 234, 204–216.
53. C. Cox and M. Ghosh, Water Res., 1994, 28, 1181–1188.
54. J. Zheng, X. Xu and D. G. Truhlar, Theor. Chem. Acc., 2011, 128, 295–305.
55. S. Suzuki, L. L. Arnold, K. L. Pennington, B. Chen, H. Naranmandura, X. C. Le and S. M. Cohen, Toxicol. Appl. Pharmacol., 2010, 244, 99–105.
56. R. Raml, A. Rumpler, W. Goessler, M. Vahter, L. Li, T. Ochi and K. A. Francesconi, Toxicol. Appl. Pharmacol., 2007, 222, 374–380.
57. T. Ochi, K. Kita, T. Suzuki, A. Rumpler, W. Goessler and K. A. Francesconi, Toxicol. Appl. Pharmacol., 2008, 228, 59–67.
58. A. Raab, S. H. Wright, M. Jaspars, A. A. Meharg and J. Feldmann, Angew. Chem., Int. Ed., 2007, 46, 2594–2597.
59. D. Mohan and C. U. Pittman, J. Hazard. Mater., 2007, 142, 1–53.
60. J. F. Ferguson and J. Gavis, Water Res., 1972, 6, 1259–1274.
61. M. Zhang, G. He and G. Pan, Chemosphere, 2015, 122, 199–205.
62. G. He, M. Zhang and G. Pan, J. Phys. Chem. C, 2009, 113, 21679–21686.
63. D. M. Sherman and S. R. Randall, Geochim. Cosmochim. Acta, 2003, 67, 4223–4230.
64. A. Adamescu, I. Hamilton and H. A. Al-Abadleh, Environ. Sci. Technol., 2011, 45, 10438–10444.
65. I. Shimoyama and Y. Baba, Carbon, 2016, 98, 115–125.
66. C. F. Matta and R. J. Boyd, in The Quantum Theory of Atoms in Molecules, Wiley-VCH Verlag GmbH & Co. KGaA, 2007, pp. 1–34.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra03813a

---