Environ. Sci. Technol. 2004, 38, 5433-5442
Thermodynamics of Nitroaromatic Compound Adsorption from Water by Smectite Clay H U I L I , † B R I A N J . T E P P E N , * ,† CLIFF T. JOHNSTON,‡ AND STEPHEN A. BOYD† Department of Crop and Soil Sciences, and Environmental Science and Policy Program, 283 Plant and Soil Sciences Building, Michigan State University, East Lansing, Michigan 48824-1325, and Crop, Soil and Environmental Sciences, Department of Agronomy, Purdue University, West Lafayette, Indiana 47907
Nitroaromatic compounds enter the environment through their use as explosives, pesticides, solvents, and synthetic intermediates in the manufacturing of dyes, perfumes, and drugs. Recent studies have found that many nitroaromatic compounds are strongly retained by smectites, especially K+-saturated smectites. Sorption occurs when nitroaromatic compounds replace water associated with the clay and form complexes between K+ and -NO2 groups. This study seeks to further understand nitroaromatic-clay interactions from the viewpoint of energetics. Adsorption isotherms of 1,3-dinitrobenzene, 1,4-dinitrobenzene, and 1,3,5trinitrobenzene from aqueous solution by K+- and Ca2+saturated smectite (SWy-2) were measured at several temperatures between 4 °C and 37 °C to determine the molar differential adsorption enthalpies. Adsorption was found to be an exothermic process on both homoionic K+- and Ca2+smectite. The smaller adsorption enthalpy on Ca-SWy-2 was consistent with its much smaller adsorption capacity for nitroaromatics compared to K-SWy-2. Our best estimate for the enthalpy of 1,3,5-trinitrobenzene interactions with K-SWy-2 is -124 kJ/mol, which is referenced to gasphase 1,3,5-trinitrobenzene, corrected for the displacement of interlayer water, and can be directly compared with quantum chemical enthalpies from the literature. Our comparable estimates for 1,3- and 1,4-dinitrobenzene interaction enthalpies are near -90 kJ/mol. We conclude that our adsorption enthalpy results are consistent with the hypothesis that nitroaromatic compounds are sorbed strongly by K-smectites because they form inner- and/or outersphere complexes with K+ cations in clay interlayers. Indeed, the basal spacings of rewetted clay films in the presence of nitroaromatic compounds imply that water molecules cannot effectively compete with the adsorbed nitrobenzenes for reactive sites on K-SWy-2.
Introduction Nitroaromatic compounds (NACs) can be released into the environment from the application of pesticides, the use of * Corresponding author phone: (517)355-0271 ext. 254; fax: (517)355-0270; e-mail:
[email protected]. † Michigan State University. ‡ Purdue University. 10.1021/es035054y CCC: $27.50 Published on Web 09/18/2004
2004 American Chemical Society
explosives, and discharge of nitroaromatic solvents and synthetic intermediates in the manufacturing of dyes, perfumes, and drugs. Such NACs are commonly found in soil and subsurface environments and pose risks to the health of humans and many other organisms (1, 2). Calculating organic matter-normalized sorption coefficients (KOM) from measured soil-water distribution coefficients (Kd) (i.e., KOM ) Kd/fOM) for NACs ignores the contribution of soil mineral fractions. The predictive utility of KOM values rests with the assumption that soil organic matter is the singular sorptive domain. However, for nitroaromatics, it is clear that other soil components (i.e., mineral fractions) may contribute to the retention of these compounds in soils and sediments. Recent studies indicate that such compounds are strongly retained by clay minerals, especially low-charged smectites saturated with weakly hydrated exchangeable cations (e.g., K+, NH4+) (3-6) as well as smectite-rich soils (our unpublished data). Several recent studies have provided insight into the operative interaction mechanisms of the adsorption of NACs by clay surfaces (5-8). Combining the results from macroscopic sorption measurements, Fourier transform infrared (FTIR) spectroscopy, X-ray diffraction (XRD), and molecular dynamics simulations as well as quantum chemistry, it was concluded that the -NO2 groups can effectively form innerand/or outer-sphere complexes with the weakly hydrated exchangeable cation (i.e., K+), whereas the tightly bound water molecules surrounding more strongly hydrated cations (e.g., Ca2+, Mg2+, Ba2+) inhibit the formation of such complexes. While the connection between FTIR studies of clay films and adsorption from water can be tenuous due to hydration effects, we demonstrated a solid link between our FTIR results and macroscopic sorption data (7, 9). We found that adsorption of 1,3,5-trinitrobenzene on K+-smectites manifested shifts in the -NO vibrational bands, while the band positions remained constant when adsorbed by Mg2+-, Ca2+-, Ba2+-smectites (4, 6, 7, 9). The aromatic ring in nitroaromatics orients parallel to clay siloxane sheets, enhancing the stability of the complexes via nonspecific van der Waals interactions with the hydrophobic portion of clay surfaces and minimizing its contact with water (6, 9, 10). The comparatively smaller hydration sphere around K+ (as compared with Ca2+) leaves a greater portion of clay surfaces available for hydrophobic adsorption. Similarly, lower chargedensity clays have larger exposed areas between exchangeable cations on the neutral siloxane surfaces, which contribute more favorable adsorption domains manifesting a greater adsorption (4, 5). Haderlein and co-workers proposed an electron donor-acceptor (EDA) mechanism to account for the strong affinity of nitroaromatics on aluminosilicate clays (5, 8). According to this mechanism, the electron-deficient aromatic π-system of nitroaromatics (due to electron withdrawal by -NO2 groups) accepts electrons from siloxane oxygens with negative-charge character (due to isomorphic substitution) to create the proposed EDA complexes. However, quantum calculations (6, 10) and FTIR results (7, 9) seem to explain sorption by complexation without need for the EDA interaction mechanism. In the study described here, we seek to improve our understanding of nitroaromatic-clay interactions in aqueous systems from the standpoint of reaction energetics. Adsorption isotherms of 1,3-dinitrobenzene, 1,4-dinitrobenzene, and 1,3,5-trinitrobenzene from aqueous solution by K+- and Ca2+saturated smectites (SWy-2) were measured at several temperatures between 4 °C and 37 °C. The temperaturedependent sorption isotherm parameters were then used to VOL. 38, NO. 20, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5433
calculate the molar differential adsorption enthalpies. The adsorption process involves combined processes of solute removal from the aqueous phase along with concurrent formation of new interactions with hydrated clay surfaces. The enthalpy of solute hydration can be calculated using solubility data measured at various temperatures (again, 4 to 37 °C). The enthalpy of nitroaromatic-clay surface interactions was then obtained from the observed adsorption enthalpy by subtracting the enthalpy for partial dehydration of the solute. These interaction enthalpies were examined for clues to further elucidate the mechanistic interactions between sorbates and clay mineral sorptive sites.
Thermodynamics of Adsorption Enthalpy Sorption energies have been widely utilized to investigate the adsorption/partitioning of aqueous phase organic compounds to natural sorbents (11-24). Sorption enthalpy, which includes the contributions of both (1) solute sorption by the hydrated solid interfacial phase and (2) removal from the aqueous phase, is most frequently estimated from sorption isotherm data at different temperatures. Isosteric heat is commonly used in evaluating the energetics of sorption for nonlinear sorption isotherms (12, 13, 19, 23). Isosteric sorption heat is the differential enthalpy change for transferring one mole of solute from the aqueous phase to the sorbent, assuming an infinite sorbent mass and solution volume at a given adsorbed concentration. However, Hill (25) pointed out that the isosteric heat does not represent the true molar differential adsorption enthalpy: He showed that the isosteric heat is equivalent to the molar differential adsorption heat plus a term describing the change of sorbent surface spreading pressure vs the change of temperature. Based on this work, Mills and Biggar (16) calculated the adsorption enthalpies of γ- and β-lindane by clay and soils from aqueous solution at a constant spreading pressure and compared them with the isosteric heats. They noted that isosteric heats increased with increasing spreading pressure, whereas the differential sorption enthalpy decreased. At equilibrium the chemical potentials of solute are equal in the bulk solution and on the solid phase, thereby the solute fugacities (f) in bulk solution (fw) and on the adsorbent (fad) are equal as well. The solute fugacities are a function of pressure (P), temperature (T), solute concentrations in the aqueous phase (Cw) and on the sorbent (Qad), as well as the two-dimensional spreading pressure (π) on the sorbent surfaces and can be expressed as
ln fw ) f(P,T,Cw)
(1)
ln fad ) f′(P,T,Qad,π)
(2)
then
d(lnfw) )
( ) ∂lnfw ∂P
dP + T,Cw
( )
∂lnfad d(lnfad) ) ∂P
( ) ∂lnfw ∂T
dT +
P,Cw
( )
( ) ∂lnfw ∂Cw
T,P
dCw
( )
in which
( ) ∂lnfw ∂T
5434
9
P,Cw
)-
∆Hhyd RT2
∂lnfad ∂T
)-
∆Hint
(6)
RT2
P,Qad,π
where ∆Hhyd and ∆Hint are the solute enthalpy changes associated with hydration in solution and with adsorbateadsorbent interactions, respectively, and R is the universal gas constant. Following Mills and Biggar (15, 16) to assume an idealdilution condition for the solute in both solution and solid phases, Henry’s law can be applied:
( ) ( ) ( ) ( ) ( ) ( ) ∂lnfw ∂Cw
∂lnfad ∂Qad
∂lnCw ∂Cw
)
T,P
∂lnQad ∂Qad
)
P,T,π
)
T,P
)
P,T,π
1 Cw
(7)
T,P
1 Qad
(8)
P,T,π
At constant P and π, we combine eq 5, eq 7 with eq 3, and eq 6, eq 8 with eq 4. The results are substituted into d(lnfw) ) d(lnfad) and yield
( )
Qad ∆Hint - ∆Hhyd Cw )1 R d T
dln
()
(9)
Equation 9 is similar to the Clausius-Clapeyron equation but with the values of Qad and Cw at constant π that can be obtained using sorption isotherms and the Gibbs surfacetension equation. For linear sorption isotherms across different temperatures, the values of Qad/Cw are constant, thereby leading to a constant observed molar adsorption heat (∆Hobs) that is equal to ∆Hint - ∆Hhyd. For nonlinear sorption isotherms, the molar adsorption heat is not a constant but depends on the adsorbate loadings on sorbent surfaces. The Gibbs surface-tension equation can be expressed as
dγ ) - RTΓsdlnCw
(10)
in which Γs is the solute surface excess and equal to Qad/S (S is the adsorbent specific surface area). For linear sorption isotherms (Qad ) KdCw), the twodimensional spreading pressure is defined as
π)
∫
γ0
γ
dγ ) - RT
∫
0
Cw
Kd Qad dlnCw ) RT Cw S S
(11)
which demonstrates that the π value is linearly proportional to the solute concentration in the aqueous solution. For a nonlinear sorption isotherm that is fit to the Freundlich equation (Qad ) KFCN w, where KF is the Freundlich sorption coefficient and N is a describer of isotherm nonlinearity), the two-dimensional spreading pressure on the solid surfaces is
(3)
∂lnfad dP + dT + ∂T P,Qad,π T,Qad,π ∂lnfad ∂lnfad dQ + dπ (4) ∂Qad P,T,π ad ∂π P,T,Qad
( )
( )
(5)
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 20, 2004
π)
∫
γ0
γ
dγ ) - RT
KFCN Qad w dlnCw ) RT Cw S NS
∫
0
(12)
If a nonlinear isotherm is described by the Langmuir equation (Qad ) KQ0adCw/(1 + KCw), where K is a constant related to interaction strength and Q0ad is the maximum adsorption for a monolayer coverage), then the surface spreading pressure is
Q0adln(1 + KCw) S
π ) RT
(13)
TABLE 1. Selected Chemical and Physical Properties of Nitroaromatic Compounds Used in This Study Sw molecular melting point (µmol/mL) weight (°C)a (g/mol)a (25 °C)a
chemicals 1,3,5-trinitrobenzene 1,3-dinitrobenzene 1,4-dinitrobenzene a
Data from ref 37.
b
213.12 168.11 168.11
122 89.8 173
1.81 3.17b 0.41b
log Kow a
log Koc a
1.18 1.30 1.49 1.56 1.46/1.49 1.46
Data from ref 38.
At a given surface spreading pressure, the corresponding solute concentration in the aqueous phase can be calculated using eq 11, eq 12, or eq 13 and substituted into eq 9 along with the associated Qad to estimate the apparent molar differential adsorption enthalpy.
Materials and Methods Sorbent. The smectite clay (SWy-2) was purchased from the Source Clays Repository of the Clay Minerals Society at Purdue University (West Lafayette, IN). The SWy-2 has a cation exchange capacity of 82 cmolc/kg and a surface area of 750 m2/g (4). The 97%, and 1,3,5-trinitrobenzene was purchased from ChemService Inc. (West Chestnut, PA) with a reported purity of 99%. These compounds were used as received. Selected physical and chemical properties of these compounds are given in Table 1. Calcium chloride dihydrate (>99%), potassium chloride (>99%), and methanol (HPLC grade) used in this study were purchased from Mallinckrodt Baker, Inc. (Phillipsburg, NJ). Solubility Measurements. The solubilities of the NACs in the aqueous phase were measured using a bottle-shaking method. A small excess of solute was added into centrifuge tubes containing Milli-Q water, and the tubes were then mixed end-over-end (40 rpm) in incubation rooms at 4.5, 13.0, 24.5, 30.0, and 37.0 °C with maximum temperature variations of (0.2 °C. At selected times during 1-week contact periods, tubes were centrifuged at 5880g for 30 min with the centrifuge temperature setting at the incubation temperature. An aliquot of supernatant was diluted to the linear range of the compound’s calibration curve and analyzed by high performance liquid chromatography (HPLC). The results showed that the solubility equilibrium was approached in 3 to 4 days. All samples were performed in triplicate. Sorption Isotherm Measurements. The sorption isotherms of NACs from an aqueous 0.01 M KCl solution for K-SWy-2 and from an aqueous 0.005 M CaCl2 solution for Ca-SWy-2 were measured using a batch equilibration method. A series of initial solute concentrations were prepared for each individual compound. Clay mass-to-solution volume ratios were adjusted to achieve sorption of 20-80% of solute from initial solution but were fixed for a given sorption isotherm. Clays were weighed into Type I, Class B Borosilicate Kimble glass centrifuge tubes, solute solutions were added, and the tubes were closed with Teflon-lined screw caps. The tubes were then shaken end-over-end at 40 rpm for 3 days at 4.5, 13.0, 24.5, 30.0, and 37.0 °C in the incubation rooms with maximum temperature variations (0.2 °C, followed by
centrifugation at 4300g for 30 min at the same temperature as the shaking temperature. Several previous studies have shown that the adsorption of NACs by clays from aqueous solution reach apparent equilibrium within a few minutes (3-5). An aliquot of supernatant was transferred to a borosilicate autosampler vial for HPLC analysis. All samples were prepared in duplicate. Supernatants were assayed for solute concentration using a Perkin-Elmer HPLC system (Binary LC pump 250 with a Series 200 autosampler) equipped with a Perkin-Elmer UVvisible detector (Series 200) and a Superco ABZ+ column (15 cm by 4.6 mm i.d.). The UV absorption wavelength was 265 nm, and the mobile phase was set as 55/45 (volume ratio) methanol/water with a flow rate of 1.0 mL/min. Controls consisted of the combinations of clay and electrolyte solution, each initial solute concentration in the supporting electrolyte, and blanks of electrolyte solution. No changes in solute concentrations were detected in the tubes containing only initial solutions, and spectroscopic studies (7) of these three solutes interacting with similar smectites showed no evidence of NAC degradation, so all solute not detected in the supernatant of each clay slurry was assumed to be sorbed by the clay. The sorbed concentrations were calculated from the difference between the initial and equilibrium solute concentrations in aqueous solution. X-ray Diffraction Analysis. After the supernatant sample was collected, the remaining solution was removed leaving approximately 2 to 3 mL residue in the tubes. The clay slurry was resuspended and then dropped on a glass slide using a disposable glass pipet. The clay suspensions were air-dried to obtain oriented films that were subject to X-ray diffraction (XRD) analysis. XRD spectra of clay films were obtained using a Philips APD 3720 automated X-ray diffractometer equipped with Cu-KR radiation, an APD 3521 goniometer, and a diffracted-beam monochromator. The scanning angle (2θ) ranged from 3 to 15° at steps of 0.02°, and the scanning time was 2 s per step. Selected air-dried clay films were subsequently exposed to 100% relative humidity (RH) for one week, and then their XRD patterns were recorded again with water present in the sample holder chamber of the X-ray diffractometer to maintain 100% RH. The basal spacing of each clay film was calculated using Bragg’s Law and the angle 2θ corresponding to the centroid of the main diffraction peak.
Results Adsorption Enthalpies on K-SWy-2. Adsorption isotherms of 1,3-dinitrobenzene, 1,4-dinitrobenzene, and 1,3,5-trinitrobenzene by K-SWy-2 at the temperatures of 4.5, 13.0, 24.5, 30.0, and 37.0 °C are shown in Figure 1. All sorption isotherms were nonlinear with curvatures concave to the abscissa. The magnitude of sorption at a given solute aqueous concentration followed the order 1,3,5-trinitrobenzene > 1,3-dinitrobenzene ≈ 1,4-dinitrobenzene. The sorption of all three nitrobenzenes on K-SWy-2 decreased with increasing temperature, demonstrating that the NAC adsorption is an exothermic process. The observed molar differential enthalpies of adsorption on K-SWy-2 were calculated using eqs 9 and 13. To use these equations to compute Qad and Cw under the constraint of constant swelling pressure across temperatures, each data set (Figure 1) needed to be fit to a continuous function. Here, the function chosen has no mechanistic implication but simply needs to fit the Qad versus Cw data very well. For this purpose, the sorption isotherms for 1,3-dinitrobenzene and 1,4-dinitrobenzene were fit to the Langmuir equation, and the 1,3,5-trinitrobenzene data were fit to the two-site Langmuir equation. The latter function was chosen only because the one-site Langmuir gave an inadequate fit to the data for 1,3,5-trinitrobenzene. Note that fitting the sorption isotherms to any other continuous function that fully VOL. 38, NO. 20, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5435
FIGURE 1. Adsorption isotherms of (A) 1,3,5-trinitrobenzene, (B) 1,3-dinitrobenzene, and (C) 1,4-dinitrobenzene on K-SWy-2 measured at several temperatures.
TABLE 2. Sorption Isotherm Parameters for Nitroaromatic Compounds on K-SWy-2 and the Range of Observed Adsorption Enthalpiesc isotherm parameters chemicals 1,3,5-trinitrobenzenea
1,3-dinitrobenzeneb
1,4-dinitrobenzeneb
temperature (°C) 4.5 24.5 30.0 37.0 4.5 13.0 24.5 30.0 37.0 4.5 13.0 24.5 30.0 37.0
Qad0 (µmol/g)
K (mL/µmol)
0.233 0.252 0.244 0.255 0.266 0.267 0.293 0.288 0.285 0.293 0.304 0.293 0.309 0.310
7.40 × 2.43 × 105 1.89 × 105 1.16 × 105 9.11 × 103 7.06 × 103 4.22 × 103 3.50 × 103 2.96 × 103 8.88 × 103 6.19 × 103 4.94 × 103 4.05 × 103 3.38 × 103
Qad0′ (µmol/g)
K′ (mL/µmol)
∆Hobs (kJ/mol)
0.142 0.181 0.348 0.245
1.40 × 2.97 × 103 1.07 × 103 1.15 × 103
-31.9 - -23.4
105
a Fitted with the two-site Langmuir equation (Q ) Q0 KC /(1 + KC ) + Q0′ K′C /(1 + K′C )). ad w w w w ad ad + KCw)). c For all curve fits, R2 > 0.99 (see Figure 1).
described the data would have worked just as well. The resultant optimal fittings are shown as solid curves in Figure 1, and the isotherm parameters are listed in Table 2. The fitting parameters for the sorption isotherm at 4.5 °C, and the adsorbent specific surface area were used in eq 13 to calculate the π values, which were then applied to eq 13 in estimating the corresponding solute concentrations in the aqueous solution at other temperatures. The values of ln(Qad/Cw) at a given π were plotted and linearly fit against 1/T, and the slope of the linear regression -∆Hobs/R was computed. The ln(Qad/Cw) was replaced with ln(KQ0ad/(1 + KCw)) for the Langmuir equation-fitted isotherms with a representative example shown in Figure 2A and with ln(KQ0ad/(1 5436
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 20, 2004
104
-23.0 - -19.6
-19.0 - -17.4
b
0 Fitted with the Langmuir equation (Qad ) Qad KCw/(1
+ KCw) + K′Q0′ ad/(1 + K′Cw)) for the two-site Langmuir equation-fitted isotherms, respectively. The values for estimated ∆Hobs as a function of the range of solute sorbed concentrations at 24.5 °C are shown in Figure 2B-D. The release of heat (i.e., more negative ∆Hobs) progressively diminished with increasing solute loading, suggesting the clay surfaces are heterogeneous with more energetic interactions occurring at lower sorbed concentrations. Alternatively, steric crowding at higher adsorbed concentrations may force the NACs into less optimal interactions with cation and/or clay surfaces, resulting in less favorable enthalpies. The magnitude of apparent sorptionenthalpy follows the
0 FIGURE 2. (A) The relationship between ln(KQad /(1 + KCw)) and 1/T as exemplified with 1,3-dinitrobenzene sorption on K-SWy-2 at several sorbed concentrations at 24.5 °C. The apparent adsorption enthalpies of (B) 1,3,5-trinitrobenzene, (C) 1,3-dinitrobenzene, and (D) 1,4dinitrobenzene on K-SWy-2 as a function of the adsorbed concentrations at 24.5 °C.
same trend as the magnitude of sorption: 1,3,5-trinitrobenzene > 1,3-dinitrobenzene > 1,4-dinitrobenzene. Adsorption Enthalpies on Ca-SWy-2. The adsorption isotherms of the three nitrobenzenes by Ca-SWy-2 at the temperatures of 4.5, 24.5, 30.0, and 37.0 °C were measured and are shown in Figure 3. The sorption manifested linear isotherms and the sorption coefficients (slope from linear fitting) are reported in Table 3. The sorption of 1,3,5trinitrobenzene is a factor of approximately 4.6 times greater than that of 1,3-dinitrobenzene and ∼4.0 times greater than that of 1,4-dintrobenzene by Ca-SWy-2. Like the adsorption on K-SWy-2, sorption on Ca-SWy-2 is also an exothermic process, as evident from decreasing solute adsorption with increasing temperature. The molar differential adsorption heat was calculated using eq 9, and the results are listed in Table 3. For a linear sorption isotherm, the sorption coefficient (Qad/Cw) is a constant hence ∆Hobs is independent of the solute loadings. The apparent adsorption heats are very similar in magnitude (about -10 kJ/mol) for all three NACs studied. Dissolution Enthalpies. The measured aqueous solubilities of 1,3- and 1,4-dinitrobenzene and 1,3,5-trinitrobenzene at 4.5, 13.0, 24.5, 30.0, and 37.0 °C are listed in Table 4. The magnitude of aqueous solubilities follows the order: 1,3dinitrobenzene > 1,3,5-trinitrobenzene > 1,4-dinitrobenzene at all temperatures studied. The reduced solubilities with decreasing temperature imply that the dissolution of these nitrobenzenes into the water phase is an endothermic process that was further quantified using the following equation
∆Hdis d(lnX) )1 R d T
()
(14)
where X is the molar fraction solubility of nitroaromatic compounds, and ∆Hdis is the enthalpy of solute dissolution in aqueous phase from the solid state. The values of ∆Hdis were obtained from the slopes of linear relationships of ln X against 1/T (Figure 4). The solubility data at 25.0 °C reported in the literature (Table 1) were also included in Figure 4. The resultant ∆Hdis values are 23.0, 26.8, and 24.5 kJ/mol for 1,3,5trinitrobenzene, 1,3-dinitrobenzene, and 1,4-dinitrobenzene, respectively. X-ray Diffraction Analysis. The measured basal spacings of the air-dried and rewetted K-SWy-2 clay films were plotted against their corresponding concentrations of adsorbed NACs (Figure 5). The data for 4,6-dinitro-o-cresol were adapted from a previous study (26) for comparison. Basal spacings of air-dried films increased progressively from 10.2 to 12.2 Å as the sorbed NAC concentrations increased, demonstrating the intercalation of NACs into the smectite interlayers. At lower loadings (i.e., 0 to 70 µmol/g sorbed concentration) the K-SWy-2 basal spacing expanded from 10.2 to 10.6 Å for 1,4-dinitrobenzene and 1,3-dinitrobenzene and to 11.3 Å for 4,6-dinitro-o-cresol. At higher loadings (i.e., 70 to 200 µmol/g sorbed concentration) the basal spacings sharply expanded to 12.1 Å. The lower basal spacings at lower loadings for 1,3-dinitrobenzene compared with 4,6-dinitro-o-cresol and 1,3,5-trinitrobenzene may be due to the larger molecular sizes of the latter two compounds. With larger NAC adsorption (i.e., >200 µmol/g sorbed concentration), the basal spacing only expanded from 12.1 to 12.2 Å, implying that 12.2 Å is a nearly optimal spacing. Typical nitroaromatic molecules are about 3 Å thick (4, 6), and dioctahedral clay layers with no species intercalated between them (i.e., pyrophyllite) have basal spacings of 9.2 Å, so the 12.2-Å spacing allows the VOL. 38, NO. 20, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5437
TABLE 4. Measured Aqueous Solubilities of Nitroaromatic Compounds at Several Temperatures aqueous solubility (µmol/mL) chemicals 1,3,5-trinitrobenzene 1,3-dinitrobenzene 1,4-dinitrobenzene a
4.5 °C
13.0 °C
24.5 °C
30.0 °C
37.0 °C
0.875 ND 1.716 2.011 2.490 (0.013)a (0.047) (0.043) (0.064) 1.526 1.797 3.025 3.815 4.888 (0.048) (0.025) (0.055) (0.063) (0.094) 0.180 0.242 0.347 0.440 0.533 (0.008) (0.026) (0.094) (0.025) (0.008)
Standard deviations are reported in parentheses.
FIGURE 4. Aqueous molar fraction solubility of nitroaromatic compounds investigated as a function of temperature. The slope of each linear fitting gives the value of (-∆Hdis/R).
FIGURE 3. Adsorption isotherms of (A) 1,3,5-trinitrobenzene, (B) 1,3-dinitrobenzene, and (C) 1,4-dinitrobenzene on Ca-SWy-2 measured at several temperatures.
TABLE 3. Sorption Coefficients of Nitroaromatic Compounds on Ca-SWy-2 at Several Temperatures and the Adsorption Enthalpies Derived from Themb chemicals
sorption coefficients Kd (mL/g)a ∆Hobs 4.5 °C 24.5 °C 30.0 °C 37.0 °C (kJ/mol)
1,3,5-trinitrobenzene 67.90 1,3-dinitrobenzene 14.72 1,4-dinitrobenzene 17.43
53.27 11.23 13.31
48.42 10.58 12.31
42.27 9.03 10.63
a Fitted with the linear sorption equation Q ad ) KdCw. fits, R2 > 0.99 (see Figure 3).
b
-10.0 -9.9 -10.5
For all curve
nitroaromatic molecules to exactly fit into the interlayer and simultaneously interact with both opposing siloxane surfaces. Exposure of previously air-dried K-SWy-2 to 100% RH resulted in the swelling of basal spacing from 10.2 to ∼15.0 Å in the absence of adsorbed NACs. The presence of lower solute loadings (i.e., < 10 µmol/g sorbed concentration for 1,4-dinitrobenzene) did not cause a change of interlayer distance compared with that in the absence of NACs. However, the presence of NACs at higher concentrations constrained the expansion of rewetted clay films. Expansion of the rewetted clay films was reduced from ∼15.0 to 12.2 Å as the solute loadings increased from 10 to 70 µmol/g and maintained the 12.2 Å (with a variation range < 0.2 Å) spacing at higher solute loadings. That is, moderate loading of NACs into the clay interlayer prevented smectite swelling in an atmosphere saturated with water. As for air-dried Ca-SWy-2 5438
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 20, 2004
FIGURE 5. The basal spacings of air-dried and rewetted K-SWy-2 films as a function of nitroaromatic compound loadings. films, the adsorption of NACs did not alter expansion relative to solute-free sample; in all cases a basal spacing at ∼15.1 Å was observed. Exposure to 100% RH only resulted in the increase of basal spacing to ∼15.5 Å for all clay films with or without adsorbed NACs.
Discussion Enthalpies of Interaction between Nitroaromatics and K-SWy-2. The adsorption of NACs to mineral surfaces from an aqueous phase involves removal of the organic solute from solution and the inception of new solute-adsorbent interactions. The observed adsorption enthalpy changes ∆Hobs are relative to the initial solution-phase state, so may be expressed as
∆Hobs ) ∆Hint - ∆Hhyd
(15)
∆Hint ) ∆Hobs + ∆Hhyd
(16)
so that
The purest measure for hydration of a compound is normally defined (27) as transfer of the compound from its ideal gas (where it has no energetic interactions with its environment) to its aqueous solution. For example
1,3,5-trinitrobenzene (g) T 1,3,5-trinitrobenzene (aq) (17) Note that this reaction is the sum of the dissolution reaction
1,3,5-trinitrobenzene (s) T 1,3,5-trinitrobenzene (aq) (18) and the reaction
TABLE 5. Enthalpy Component Analysis for Nitroaromatic Compounds Interacting with Aqueous Solutions and K-Clay Surfacesj compound
∆Hdisa ∆Hsubl ∆Hhyde ∆Hobsf ∆Hintg ∆Hinth ∆Hinti
1,3,5-trinitrobenzene 23.0 107b 1,3-dinitrobenzene 26.8 87c 1,4-dinitrobenzene 24.5 95d
-84 -60 -70
-32 -116 -99 -124 -21 -81 -69 -89 -17 -87 -73 -93
a This work. b Reference 28. c Reference 29. d Mean of 96.2 (29) and 94.3 (30). e Computed using eq 20. f This work, using values at loading rates of 150 µmol/g on K-SWy-2 (extrapolated for 1,4-dinitrobenzene). g Assuming complete dehydration of the nitroaromatics (i.e., no interlayer water), using eq 16. h Assuming 80% dehydration of the nitroaromatics within the clay interlayer, using eq 21. i Assuming 80% dehydration of the nitroaromatics within the clay interlayer and correcting the result for the displacement of approximately 10 (for 1,3,5trinitrobenzene) and eight (for 1,3- and 1,4-dinitrobenzene) water molecules from the smectite interlayer. j Enthalpies are reported for dissolution (∆Hdis), sublimation (∆Hsubl), hydration (∆Hhyd) referenced to the gas phase, adsorption from aqueous solution (∆Hobs), and adsorption (∆Hint) referenced to the gas phase, all in kJ mol-1.
1,3,5-trinitrobenzene (g) T 1,3,5-trinitrobenzene (s) (19)
of the nitroaromatic compound in such a system to be roughly 80%, which would result in modification of eq 16 to
Reaction 19 is the reverse of the sublimation reaction, so the hydration enthalpy can be expressed as (27)
∆Hint ) ∆Hobs + 0.8‚∆Hhyd
∆Hhyd ) ∆Hdis - ∆Hsubl
(20)
Fortunately, estimates for sublimation enthalpies ∆Hsubl have been published (28-30) for the NACs studied here and are listed in Table 5. The resulting hydration enthalpies for our compounds, computed using eq 20, seem rather large in magnitude (Table 5). Are they reasonable? We could find no direct determinations of pure hydration enthalpies for these NACs in the literature, but we can crudely estimate them: The best estimate for the hydration enthalpy of benzene is -31.7 kJ/mol (27). The effect of adding a -NO2 group can be estimated from homologous compounds. Gibbs free energies of hydration are much more common in the literature than enthalpies, and most of the free energy change can plausibly be ascribed to enthalpy since adding a -NO2 group does not significantly alter the size and hence the entropy of hydration. When a single -NO2 group is added to ethane, propane, butane, benzene, toluene, and phenol, the change in the free energy of hydration is -18 ( 5 kJ/mol ((31): If just benzene, toluene, and phenol are considered, then the change in the free energy of hydration is -14 ( 3 kJ/mol). Using -18 kJ/mol as our estimate for the hydration enthalpy change per -NO2 group, and adding multiples of this change to the benzene enthalpy of hydration, we get -68 kJ/mol for the hydration enthalpy of dinitrobenzenes and -86 kJ/mol for 1,3,5-trinitrobenzene. These numbers are in fair agreement with our estimates (obtained using eq 20) in Table 5 for ∆Hhyd of the NACs. Table 5 also shows three sets of estimates for ∆Hint, the net enthalpies of interaction between the nitroaromatic solutes (referenced to the gas phase) and hydrated K-SWy-2, found using eq 16. Note that the full value of ∆Hhyd in eq 16 can only be used if the solute is completely removed from aqueous solution. The values of ∆Hint calculated under this assumption are shown but probably are not accurate since we do not believe the clay interlayers are fully dehydrated. In fact, Figure 5 shows that, at loading rates less than 200 µmol/g, K-SWy-2 will adsorb water even from a watersaturated atmosphere. Molecular dynamics simulations of nitroaromatics sorbed to K-SWy-2 at realistic loading rates and d spacings of 12.3 Å indicate that a few water molecules interact with the edges of each nitroaromatic molecule, but no water interacts with either aromatic planar surface of the molecule (4). Thus, we estimate the degree of dehydration
(21)
Application of eq 21 results in estimates for ∆Hint of -99, -69, and -73 kJ/mol for 1,3,5-trinitrobenzene, 1,3-dinitrobenzene, and 1,4-dinitrobenzene, respectively (Table 5), relative to their gas-phase energies. Nitroaromatics on K-SWy-2 manifest a strong sorption and nonlinear isotherms (Figure 1), and it is clear that they occupy interlayer sites (Figure 5). Entry of NACs into the interlayer requires the NACs to replace many of the water molecules that were previously intercalated. Thus, the first two values of ∆Hint reported in Table 5 do not reflect the ‘simple’ process of direct interactions between solute and clay sorptive sites but are summations of several processes including energy-consuming steps such as water removal. The adsorption process can be conceptualized as creation of an empty cavity by removing water from within the clay layer and then filling the cavity with the organic solute (32). The first two ∆Hint reported in Table 5 are the sum of those two processes and, therefore, should underestimate the enthalpies of ‘simple’ solute-clay interaction that would be directly comparable with computational estimates (10, 33). The heat of immersion is a measure of the enthalpy change in a clay-water system arising from hydration of the clay and its interlayer cations. The heat of immersion for K-SWy-2 is between -1 and -4 kJ/mol of water for the first layer of water loading (34). Based on the density of solid 1,3,5trinitrobenzene (1.478 g/cm3 (35)), the molecular volume of 1,3,5-trinitrobenzene is 240 Å3, or eight times that of a water molecule. Since interlayer water will pack imperfectly, the 1,3,5-trinitrobenzene cavity probably needs to be at least 10 water molecules in size, so it would require 10 to 40 kJ/mol to empty that cavity or 25 kJ/mol on average. This value should then be subtracted from our previous estimate of ∆Hint (-99 kJ/mol from Table 5) to yield a revised ‘simple’ ∆Hint of -124 kJ/mol. Thus, the best estimate for our experimental enthalpy of 1,3,5-trinitrobenzene interaction with K-SWy-2, which is referenced to the gas phase, corrected for the displacement of interlayer water, and can be directly compared with quantum chemical estimates (10, 33), is about -124 kJ/mol. Comparable estimates for 1,3- and 1,4dinitrobenzene are near -90 kJ/mol (Table 5). These enthalpies for NAC-clay interactions are quite large, and it would be informative if they could be used to distinguish between various structural hypotheses for adsorption mechanisms. Our enthalpy estimates are somewhat dependent on the NAC loading rate at the clay surface (Figure 2). The data of VOL. 38, NO. 20, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
5439
Table 5 are referenced to loading rates of 150 µmol/g, so our observed ranges of ∆Hobs (Figure 2) for the dinitrobenzenes are about 2 to 3 kJ/mol more negative than the values in Table 5, while that for 1,3,5-trinitrobenzene is up to 9 kJ/mol less negative. As we argue below, most of the latter range is probably due to unfavorable interactions between 1,3,5trinitrobenzene molecules competing for the same adsorption sites, so the “noncompetitive” range of our data is only about 2 kJ/mol (∆Hobs of -32 to -30 kJ/mol, Figure 2). Thus, the loading-rate dependence, over our observed loading rates from 20 to 250 µmol/kg, is only about (2 kJ/mol for our NAC adsorption enthalpy estimates in Table 5, which is rather small relative to the magnitudes of the enthalpies themselves. For example, the loading-rate dependence of ∆Hint for 1,3,5trinitrobenzene is -123 ( 1 kJ/mol over the NAC loading range from 150 to 250 µmol/g K-SWy-2 smectite clay. The only previous estimates for NAC interaction energies that we are aware of were computational results. Haderlein et al. (3) have experimentally estimated ∼40 kJ/mol for the adsorption enthalpy (∆Hobs) of substituted nitrobenzenes by Cs-kaolinite from aqueous solution, which is greater than the ∆Hobs values measured in this study. This indicates that NACs develop a stronger affinity with Cs-clay than with K-clay. Unfortunately, they (3) did not estimate ∆Hint. Quantum mechanical calculations (10, 33) have been used to estimate ∆Hint as the energy of the NAC-clay complex minus the energies of the uncomplexed (gas-phase) NAC and clay, with corrections for basis set superposition errors. Since our ∆Hint estimates are relative to the gas phase NACs and are corrected for the displacement of interlayer water by the sorbed NACs, they are directly comparable to such computational results. In one study (10), quantum calculations of 1,3,5-trinitrobenzene interactions with one neutral 2:1 clay surface fragment yielded an interaction energy of -38 kJ/mol. While direct extrapolation is impossible, if these results were extended such that two siloxane surfaces were present, one might expect as a first approximation that the interaction enthalpy would be near -76 kJ/mol for 1,3,5-trinitrobenzene. This is only about 60% of our best estimate (Table 5) for the enthalpy of 1,3,5-trinitrobenzene interaction with K-SWy-2. In a second study (33), quantum estimates of nitrobenzene interactions with a Na-smectite fragment were -49 kJ/mol. In this study, the -NO2 group formed two H-bonds with water molecules in the first hydration sphere of Na+. If these results for nitrobenzene (33) were simply tripled such that 1,3,5-trinitrobenzene were to H-bond to six waters of hydration around multiple cations, one might anticipate the interaction enthalpy to be near -147 kJ/mol. This value is within 20% of our estimate (Table 5). Similarly, if the nitrobenzene results (33) were doubled to provide an estimate for 1,3- and 1,4-dinitrobenzene, that straightforward estimate for ∆Hint would be -98 kJ/mol, again within 10 to 20% of our estimates (Table 5). Uncertainty again arises since the quantum calculations (33) were conducted with Na+, a small clay fragment, and only two water molecules, whereas our estimates are based on an experimental K- and watersaturated smectite system. Since the quantum system (33) was quite ‘dry’, multiple hydration shells did not build up around Na+ and just two first-hydration-shell water molecules bridged between the Na+ and the nitro-group, which may have created a system fortuitously like a wet K-NAC-smectite. Furthermore, the phenyl ring of the NAC interacted with neither clay nor water in the quantum system (33), making direct connection with our experimental results tenuous. While indirect, such calculations (10, 33) are the only comparisons that exist for our ∆Hint estimates. While bulk measurements such as enthalpies cannot provide a decisive edge to any structural or mechanistic hypothesis (36), the computational estimate of -147 kJ/mol for ∆Hint of 1,3,5-trinitrobenzene, extrapolated from ni5440
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 20, 2004
trobenzene H-bonding to waters of cation hydration (33), is much closer to the experimental enthalpy (-124 kJ/mol, Table 5) than the computational estimate of -76 kJ/mol, extrapolated from van der Waals and electrostatic interactions of 1,3,5-trinitrobenzene with one clay surface in the absence of cations (10). From the correlation of these computational results with our experiments, we conclude that our adsorption enthalpy results are quite consistent with our mechanistic hypothesis (6, 7, 9) that nitroaromatic compounds are sorbed strongly by smectites because they form inner- and/or outersphere complexes with K+ cations in clay interlayers. Again, due to key differences between the experimental and quantum systems, any agreement between their enthalpy estimates may be fortuitous. It is interesting that the interaction enthalpy results (Table 5) tend to be in an approximate ratio of 3:2 for trinitrobenzene: dinitrobenzene (more precisely, 2.8:2). Thus, if one hypothesizes on the basis of FTIR spectroscopic results (6, 7, 9) that the -NO2 groups of these compounds either form H-bonds with water molecules in the first hydration sphere of K+ or form inner-sphere complexes with K+, then one could interpret the present results to show that interaction energies between nitroaromatic compounds and interlayer species are about -40 to -45 kJ/mol per -NO2 group. This is reasonably close to the previous estimate (-49 kJ/mol (33)) for the H-bonding of one -NO2 group to waters of cation hydration. We believe that direct K+-NO2 complexation in the interlayer cannot be ruled out, either, since FTIR spectroscopy shows a significant cation-dependence of -NO2 stretching frequencies. One might expect that the interaction enthalpy for such a complex would be of even larger magnitude, but this enthalpy would be tempered by inclusion of the severe energetic penalty for dehydrating the cation in order to form an inner-sphere complex with the nitroaromatic molecule. Another piece of circumstantial evidence from this study that supports direct involvement of -NO2 groups in the adsorption mechanism is the dependence of adsorption enthalpy on the loading rate (Figure 2B). Note that there is a strong inflection to less favorable adsorption enthalpy above 1,3,5-trinitrobenzene loading rates of 250 µmol/g. The interlayer region contains approximately 820 µmol/g of K+ cations, so the enthalpy inflection occurs near a 3:1 ratio of K+:1,3,5-trinitrobenzene. At higher loading rates, 1,3,5trinitrobenzene molecules would have to begin sharing cations in order to form complexes with all three of their -NO2 groups; there would presumably be an energy penalty for doing so, and a less favorable enthalpy would result. At the highest loading rate we measured in the present study (but not the highest possible), there were about 2.3 K+ per 1,3,5-trinitrobenzene. At the largest 1,3,5-trinitrobenzene loading rate we have observed (7), this ratio was very close to 2. In studying the rewetting of previously air-dried K-SWy-2 films, the interlayer spacings of the clay as a function of solute loading indicate that sorbed NACs inhibit the rehydration of the smectite. At low loading rates (i.e.,
