Chemosphere 103 (2014) 322–328

Contents lists available at ScienceDirect

Chemosphere journal homepage: www.elsevier.com/locate/chemosphere

Sorption and solubility of ofloxacin and norfloxacin in water–methanol cosolvent Hongbo Peng a, Hao Li a, Chi Wang a, Di Zhang a, Bo Pan a,⇑, Baoshan Xing b a b

Faculty of Environmental Science & Engineering, Kunming University of Science & Technology, Kunming, Yunnan 650500, China Stockbridge School of Agriculture, University of Massachusetts, Amherst, MA 01003, United States

h i g h l i g h t s  Hydrogen bond length calculation explains OFL and NOR solubility in cosolvent.  Interaction energy calculation should incorporate more real-system-conditions.  OFL solubility decreases with increased methanol volume fraction.  Solubility of NOR varied nonmonotonically with increasing methanol volume fraction.

a r t i c l e

i n f o

Article history: Received 25 June 2013 Received in revised form 23 November 2013 Accepted 1 December 2013 Available online 2 January 2014 Keywords: Antibiotics Dissolution Environmental fate Hydrogen bond Interaction energy

a b s t r a c t Prediction of the properties and behavior of antibiotics is important for their risk assessment and pollution control. Theoretical calculation was incorporated in our experimental study to investigate the sorption of ofloxacin (OFL) and norfloxacin (NOR) on carbon nanotubes and their solubilities in water, methanol, and their mixture. Sorption for OFL and NOR decreased as methanol volume fractions (fc) increased. But the log-linear cosolvency model could not be applied as a general model to describe the cosolvent effect on OFL and NOR sorption. We computed the bond lengths of possible hydrogen bonds between solute and solvent and the corresponding interaction energies using Density Functional Theory. The decreased OFL solubility with increased fc could be attributed to the generally stronger hydrogen bond between OFL and H2O than that between OFL and CH3OH. Solubility of NOR varied nonmonotonically with increasing fc, which may be understood from the stronger hydrogen bond of NOR-CH3OH than NOR-H2O at two important sites (–O18 and –O21). The interaction energies were also calculated for the solute surrounded by solvent molecules at all the possible hydrogen bond sites, but it did not match the solubility variations with fc for both chemicals. The difference between the simulated and real systems was discussed. Similar sorption but different solubility of NOR and OFL from water–methanol cosolvent suggested that sorbate–solvent interaction seems not control their sorption. Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction Solubility is an intrinsic physicochemical property of organic chemicals that influences their release, transport, absorption efficiency, and environmental behavior (Abraham and Le, 1999). Previous studies have indicated that solubility of hydrophobic organic compounds (HOCs) is an important physicochemical property determining how these compounds partition/sorb in various environmental media such as water, soil/sediment, and organisms, and thus a useful parameter for predicting HOC sorption behavior (Chiou et al., 1983). But the similar type of study is rare (if not none) for ionizable organic chemicals. ⇑ Corresponding author. Tel./fax: +86 871 65170906. E-mail address: [email protected] (B. Pan). 0045-6535/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.chemosphere.2013.12.025

It is widely accepted that the sorption mechanisms for ionizable chemicals are much more complicated than HOCs. With the understanding that the sorption of a solute on a solid particle from a solvent involves the interactions of solute–adsorbent and solute–solvent, proper separation of these interactions could greatly facilitate the sorption mechanisms. Therefore, it is useful to study both solubility and sorption in organic solvent–water cosolvents for ionizable chemicals. The most fundamental model for the dissolution of a solute in a certain solvent is that the solute molecules diffuse into solvent cavity by overcoming intermolecular interactions. The active functional groups of solute molecules interact with solvent molecules and form solvent-shells around the solute molecules. In the cosolvent system, the solvent shell acts as a compatible layer between solute and organic solvent (Gandhi and Murthy, 2012). Besides hydrophobicity and polarity,

H. Peng et al. / Chemosphere 103 (2014) 322–328

other interactions such as ion–dipole, dipole–dipole, solute–solvent and hydrogen bond could also affect solubility (Miyako et al., 2010). Among these interactions, the ability of the solvent to form hydrogen bond with solute molecule was one of the most important factors controlling the solubility of solute in solvent (Taha and Lee, 2010; Gandhi and Murthy, 2012). Hydrogen bond has been intensively investigated because the chemical solubility, partitioning and adsorption are often controlled by this mechanism (Acelas et al., 2012; Majerz, 2012). However, the direct determination of the equilibrium constants for hydrogen bond by experimental method is not practicable. Further, the hydrogen bond length and energy cannot be measured directly (Wendler et al., 2010). Fortunately, hydrogen bond can be simulated through computational methods with low human workload and cost. For example, theoretical simulation using Density Functional Theory (DFT) has been frequently employed to evaluate the intermolecular interactions (Schoone et al., 1998). Hydrogen bond can be quantitatively described by parameters of hydrogen bond lengths and angles, and interaction energies for the hydrogen bond systems. Stronger hydrogen bond is expected when the bond length is shorter and the bond angle is close to 180° (Nishio et al., 2009). Combination of theoretical simulation with lab experiment will greatly facilitate solubility and sorption studies of ionizable organic chemicals. Among ionizable chemicals, antibiotics are well-known as a class of emerging contaminants. Ofloxacin (OFL) and norfloxacin (NOR) are fluoroquinolones (FQs) antibiotics and have caused concern due to their specific therapeutic properties to human and wide applications in agricultural and veterinary treatments including treating urinary tract infections and respiratory tract infections (Picó and Andreu, 2007). These chemicals are also applied in genetic science as specific bacterial inhibitors (Ulu, 2009). The potential risks of the wide application of antibiotics have attracted a great deal of attention, such as promoting the development of drugresistant genes. OFL and NOR have similar chemical structures, but with various physical–chemical properties. For example, their solubilities in water were different over one order of magnitude. Understanding the relationship between solubility and sorption for these chemicals will greatly promote the understanding on their environmental behavior. Carbon nanotubes (CNTs) have attracted a great deal of research attention due to their wide application and potential ecological risks (Lam et al., 2006). Because of the strong interactions between CNTs and organic contaminants (Masciangioli and Zhang, 2003; Tasis et al., 2006), the released CNTs during manucfacturing activities may significantly alter the fate of organic contaminants in the environment (Hyung et al., 2006). The interaction mechanisms between CNTs and organic chemicals have been a hot topic in sorption studies. In addition, CNTs are described as a well-defined adsorbent with explicit chemical structures (Pan et al., 2010). They are good model adsorbents for sorption experiments and have been successfully applied in sorption mechanism studies (Zhang et al., 2010). We have previously used CNTs as model adsorbents for identifying the importance of hydrophobicity and molecular structure to NOR and OFL sorption (Peng et al., 2012b), and the application of log-linear cosolvency model to explain OFL sorption (Peng et al., 2012a). But none of these discussions could be applied to separate the interactions of solute–adsorbent and solute– solvent. Therefore, this work investigated both sorption and solubility variations of OFL and NOR in methanol–water cosolvent. Hydrogen bond is the center of discussion for their solubilities. Multi-walled CNTs (MWCNTs) with different functional groups will be applied as model adsorbents. This line of study will shed light on the sorption mechanisms of ionizable organic chemicals.

323

2. Methodology 2.1. Experimental method 2.1.1. Materials Graphitized (MG), carboxylized (MC) and hydroxylized (MH) MWCNTs were purchased from Chengdu Organic Chemistry Co., Chinese Academy of Sciences. According to the information provided by the manufacturer, the purities of these MWCNTs were more than 95%. The outer diameter for these three MWCNTs ranged from 8 to 15 nm. All the CNTs were characterized for their surface areas, and the values were 117 m2 g1, 164 m2 g1 and 228 m2 g1 for MG, MC, and MH, respectively. The specific surface area was calculated from N2 adsorption and desorption isotherms using Brunauer–Emmett–Teller (BET) method (Autosorb-1C, Quantachrome). All the three CNTs were characterized for elemental compositions (MicroCube, Elementar, Germany) and surface functional groups (X-ray photoelectron spectroscopy). These characterization results are listed in Table S1. Our previous study and other studies indicated that these three CNTs were good model adsorbent because of their explicit structures and representative functional groups (Zhang et al., 2010; Peng et al., 2012a). The adsorbates used in this study were OFL and NOR, which were obtained from Bio Basic Inc. All the other chemicals were higher than analytical grade (purity >99.5%). The physical–chemical properties of selected adsorbents and adsorbates were listed in Table S1 and Table 1, respectively. 2.1.2. Adsorption experiments Adsorption of OFL and NOR on CNTs were investigated at different volume fractions of methanol in methanol/water mixture solutions. The initial concentration of OFL and NOR was fixed at 10 mg L1 in this experiment. The volume ratio of water:methanol (v:v) was distributed in the range 0.0–1.0. Adsorption experiments were carried out in glass vials with Teflon-lined screw caps. Duplicate samples were done for each concentration point including blanks (without NOR or OFL). OFL and NOR were separately dissolved in electrolyte solution containing 0.02 M NaCl (background electrolyte) and 200 mg L1 NaN3 (bio-inhibitor) as stock solutions. According to preliminary studies, 1 mg of CNTs (to maintain analytical accuracy) was added into the 4 mL solutions with varied NOR/OFL concentration. The liquid:solid ratios of 4000:1 were used in the sorption experiments to ensure 20–80% adsorption. Afterward, all the vials were stored in dark and shaken in an airbath shaker at 25 °C for 7 d, which deems to reach apparent adsorption equilibrium. After equilibration, all the vials were centrifuged at 2000 rpm for 10 min and the adsorbates in the supernatants were measured. The separation of CNTs and aqueous phase was confirmed through TOC analysis. OFL and NOR were measured by HPLC (Agilent Technologies 1200) equipped with a reversedphase C8 column (5 lm, 4.6  150 mm). The UV detector of HPLC was set at 286 nm and 280 nm for OFL and NOR, respectively. The adsorbed adsorbates at each equilibrium concentration were determined by mass balance, and calculated by the difference between the applied and equilibrated aqueous phase concentration. The mobile phase was 10:90 (v:v) of acetonitrile and deionized water with 0.8% acetic acid with a flow rate of 1 mL min1. The retention time was 4.6–5.0 min. 2.1.3. Measurement of OFL and NOR solubilities The solubilities of OFL and NOR (Cs) were measured at different volume fractions of methanol (water:methanol (v:v) = 0.0–1.0). Three mg OFL and NOR were separately added into 1.5 mL glass vials with Teflon-lined screw caps. One mL of the water–methanol

324

H. Peng et al. / Chemosphere 103 (2014) 322–328

Table 1 Adsorbate properties (solubility, Cs; n-octanol–water partition coefficient, KOW). Chemical Ofloxacin

Norfloxacin

a b c d e f g

Csa (water), mg L1 b

3400 ± 141 3250d

320 ± 16b 280f

Csa (methanol), mg L1

Csa (ethanol), mg L1

b

445 ± 10

620 ± 21b

1000 ± 8b

1210 ± 19

b

KOWa

pKa1/pKa2c b

6.10/8.28

0.251 ± 0.016b 0.1g

6.23/8.55

0.446 ± 0.025 0.407e

Chemical structure

Measured at 25 °C. Standard deviation. From Park et al. (2002). From Goyne et al. (2005). From Zorita et al. (2009). From Boonsaner and Hawker (2010). From Lorphensri et al. (2006).

mixture solution was injected in each vial. All the vials were kept in dark and shaken in an air-bath shaker at 25 °C for 2 d. Subsequent 3 mg OFL and NOR were added in the vials until solid residuals were observed. Then, these vials were shaken for another 5 d until the concentration of solute in the cosolvents did not change. All the vials were then centrifuged at 2000 rpm for 10 min and the supernatants were subjected to quantification of the solute concentration. 2.2. DFT computation method In this study, the geometric optimizations, vibrational frequencies and single-point energy of the different hydrogen bonding complexes were performed using DFT method. For the DFT method, the Becke’s three-parameter hybrid exchange function combined with the Lee–Yang-Parr gradient-corrected correlation functional (B3LYP) was used (Lee et al., 1988; Becke, 1993). Previous researches have shown that this method could ensure the calculation results precision for hydrogen-bond systems (Van Bael et al., 1997; Schoone et al., 1998). We used the 6-31++G (d, p) basis set for all the calculations. Reliable results for hydrogen-bonded systems were thus obtained as other studies recommended (Hobza et al., 1995). The geometric optimizations and vibrational frequencies analysis were performed for all the possible hydrogen-bond complexes at this level. The frequency analysis could confirm the optimized geometries, and all the systems were no imaginary frequency. Interaction energies (Eint) of all the possible hydrogenbond complexes were calculated to compare the stability of the complexes. All calculations were performed using Gaussian 09 program. Interaction energies are the difference between total energy of complexes and the sum energies of the separate subunits (Efremenko and Sheintuch, 2006). It is an important coefficient to quantitatively describe the stability of the systems. The more negative interaction energies suggested more stable complexes. We calculated the Eint for solute–solvent using one solvent molecule as well as 6 (for NOR) or 7 (for OFL) solvent molecules occupying all possible hydrogen bond sites. The Eint of the complexes were calculated by using the following equation:

Eint ¼ Eðsolute-solventÞ  EðsoluteÞ  EðsolventÞ where E(solute–solvent) is the total energy of the solute interaction with solvent, E(solute) is the total energy of an isolated solute molecule, and E(solvent) is the total energy of an isolated solvent molecule. Solute was OFL or NOR, and solvent was water or methanol in this study.

3. Results and discussion 3.1. Solubility and adsorption of OFL and NOR variation with water– methanol ratio The sorption as affected by methanol was similar for OFL and NOR. The apparent sorption decreased greatly with increasing methanol volume fractions when fc 6 0.7, but did not change significantly at higher fc (>0.7) (Figs. S2B and 1B). Adsorption isotherms of OFL and NOR on three CNTs in water and methanol are also illustrated in Fig. S3, showing that sorption affinity of OFL and NOR in water was higher than in methanol. The decreased OFL and NOR sorption when the solvents were changed from water to methanol was consistent with that of sorption isotherm study. Bouchard indicated that the sorption of hydrophobic organic chemicals (HOCs) decreased with increasing methanol fraction because of the increased solubility (Bouchard, 2002). Clearly, both chemicals investigated in this study are not HOCs. The previous discussion suggested that the effects of cosolvents on sorption have been sufficiently described by the log-linear cosolvency model (Peng et al., 2012a). This model has the form of log Km = log Kd  arfc, and a is an empirical constant hypothesized to reflect cosolvent–sorbent interactions. The parameter r is an index of the solubilizing power of the cosolvent, which describes solute–cosolvent interactions. fc is the volume fraction of cosolvent. It was concluded that for OFL, solute–colsolvent (OFL-methanol) interactions were much weaker than cosolvent–sorbents (methanol-CNTs) interactions. The methanol molecules may compete with OFL for sorption sites on CNTs. This maybe the reason that the adsorption of OFL reduced in water–methanol cosolvents (Peng et al., 2012a). However, for NOR, solubility varied nonmonotonically with methanol fraction. It increased when methanol fraction increased from 0.0 to 0.6, decreased slightly with the methanol fraction increased from 0.6 to 0.8, and then increased dramatically when methanol fraction increased from 0.8 to 1.0 (Fig. 1A). Similar trend was consistently reproducible in repeated experiments (Fig. S1). This type of solubility change could not be used to investigate the sorption change with methanol volume fractions as for NOR. It is thus important to study the interactions between solute and solvents (methanol and/or water). Various methods were applied for investigating the solute–solvent and solvent–solvent interactions. In these interactions, hydrogen bond is a commonly discussed interaction mechanism. Molecular properties such as solubility, molecular conformation and aggregation are operated in large part by hydrogen bond (Biserka and Kresimir, 2008; Wendler et al.,

325

H. Peng et al. / Chemosphere 103 (2014) 322–328

8.0

2

6.0

-1

logKd (L kg )

-1

( Cs (×10 mg L

6.0

A

4.0

B

MH y = -1.937x + 5.206 2 r = 0.985

4.0 MC y = -2.029x + 4.842 r2 = 0.966

2.0

MG y = -3.042x + 5.161 r2 = 0.983

2.0 0.0 -0.2

0.0

0.2

0.4

0.6

0.8

1.0

0.0 -0.2

1.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

fc Fig. 1. NOR solubility and sorption as affected by methanol addition in aqueous phase. The solubility variation with methanol volume fractions (fc) is presented in panel A, while adsorption coefficient variation with fc is illustrated in panel B. NOR solubility measurement at different methanol fractions was replicated another time and the data are presented in Fig. S1. The comparison between NOR and OFL solubility and sorption variations with methanol fraction is presented in Fig. S2.

2010). Systematic computations for hydrogen bond were conducted to explain the OFL/NOR solubility trend in this study. 3.2. Strength of hydrogen bond at different molecular sites Hydrogen bond of OFL-solvent and NOR-solvent at different possible sites are illustrated in Fig. 2, and their bond length and

A

VII

III

angle are summarized in Table S3. Among different intermolecular hydrogen bond systems, there are no strong hydrogen bonds 0 (bond lengths: 1.2–1.5 Å A), much of them are moderate and few are weak hydrogen bonds (Table S2). This is probably why the apparent sorption of OFL and NOR was not different giving their dramatic solubility difference. One can find that the order of the intermolecular hydrogen bond length for OFL-CH3OH

II

B

VI

III

17

14 22 19

24

18

I

19

7

1

V

12 8 4

2

16 IV

3.0

15 V

22 23

I

13

11 14

7

3

VI

C

9

5

1

20

IV

26

10 6

21 2

6

17

3

5

13

15 16

3.0

4 8

12 21

25

9

11

20

23

18

10

D OFL-H2O

2.0

OFL-CH3OH

R(Ǻ)

2.0

NOR-H2O

0.0

3.0

NOR-CH3OH

1.0

1.0

O 22

N2

O 25

F14

N5

O 24

O 18

0.0 N11 3.0

E

2.0

N14

O20

O18

F17

O21

F OFL-H2O

2.0

R(Ǻ)

OFL-CH3OH NOR-H2O 1.0

1.0

0.0

NOR-CH3OH

0.0 O 22/18 N 2/14 O 25/20

F 14/17

N 5/11

O 24/21

O 22/18 N 2/14 O 25/20

F 14/17

N 5/11

O 24/21

Fig. 2. Possible molecular sites of hydrogen bond for OFL (Panel A), NOR (Panel B) and the corresponding hydrogen bond length of OFL-water/methanol (Panel C), NOR-water/ methanol (Panel D). OFL and NOR were also compared in water (Panel E) and methanol (Panel F). R: hydrogen bond length (dark gray: C; red: O; blue: N; green: F). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

326

H. Peng et al. / Chemosphere 103 (2014) 322–328

is –O22 < –N2 < –O25 < –F14 < –N5 < –O24 < –O18, which is the opposite trend for hydrogen bond strength. The order of the strength of hydrogen bond for OFL-CH3OH was the same to that of OFL-H2O (Fig. 2C). Methanol could replace the OFL-bound water at the site following the order of hydrogen bond strength when methanol was added in OFL solution. Fig. 2C illustrated that the hydrogen bond lengths of OFL-CH3OH at most sites are lower than those of OFL-CH3OH except –N5. Thus, the hydrogen bond interaction between OFL and water were higher than that of OFL and methanol. OFL solubility decreased gradually with increasing methanol fraction in our experiments, possibly due to the fact that the hydrogen bond interaction between OFL and cosolvent became lower as the methanol fractions increased. For NOR, hydrogen bond strength of NOR and methanol in –O18 (1.84 Å) site was the strongest for all the hydrogen bonds between NOR and water/methanol. The hydrogen bond between NOR and methanol became stronger as the methanol initially added. This may be the reason that NOR solubility increased with increasing methanol fraction. According to the hydrogen bond length order, methanol replaced water molecules at other sites with moderate strength when the fraction of methanol increased. The longer hydrogen bond length of N11, N14, and F17 formed between NOR and methanol in comparison to NOR and water could result in the apparent decreased solubility of NOR with increasing methanol fraction from 0.6 to 0.8. When the methanol fraction increased to 0.8, the fraction of water was relatively low in solution. The –O21 became the replacing center because of its longest hydrogen bond length. But the hydrogen bond strength of NOR and methanol in – O21 (2.291 Å) was much stronger than those of NOR and water in – O21 (2.969 Å). The increased hydrogen bond between NOR and methanol could explain the data of increased NOR solubility as the methanol fraction increased from 0.8 to 1.0. We also compared the hydrogen bond lengths between OFL and NOR in water (Fig. 2E) and methanol (Fig. 2F). It was obvious that the hydrogen bond lengths of OFL and water were shorter than those of NOR and water in most sites except –N5/11 (Fig. 2E). This O25

O24

N2

O22

Eint (KJ/mol)

0

N5

O18

F14

3.3. Interaction energy of hydration systems Dielectric constant and hydrogen bond might decrease but the specific interactions between solute and solvent molecules increased as the methanol fraction increased in water (Desai et al., F17

N11

O21

0

-10

-10

-20

-20

N14

O20

O18

OFL-H2O OFL-CH3OH NOR-H2O NOR-CH3OH

-30

-30

A

-40 O 25/20 O 24/21 N 2/14

Eint (KJ/mol)

observation indicated that the hydrogen bond between OFL and water was stronger than that of NOR and water, which probably resulted in about one order of magnitude higher solubility of OFL than that of NOR in water. Hydrogen bond lengths were similar for OFL/NOR with methanol of all sites (Fig. 2F). This maybe the reason that the solubility of OFL and NOR in methanol was similar as shown by our experimental results. The interaction energies for the hydrogen bonds between solutes and solvents at different atomic sites were also calculated using DFT. Stronger interaction is expected when the interaction energy (Eint) values are more negative. Most of the computed Eint values for OFL-water were much lower than those of OFL-methanol (Fig. 3A). The only exception is the Eint value of –O18, which is less important than the other sites because of its less negative value. This calculation suggested that OFL-water interaction was stronger than OFL-methanol, and thus, the solubility of OFL in water was higher than that of in methanol. For NOR, two (–O20 and –O18) of the most important four sites showed stronger interaction with water and two (–O21 and –N14) of them had stronger interaction with methanol (Fig. 3B). This may also partly contribute to the variation of NOR solubility change with increasing methanol fraction. The interaction energies between OFL and NOR in water (Fig. 3C) and in methanol (Fig. 3D) were also compared. The computed Eint values of OFL-water were generally lower than those of NOR-water in most sites except –O22/18 site (Fig. 3C), and thus the interaction energies of OFL-water were higher than that of NORwater. This also contributes to the observed higher solubility of OFL than NOR in water. For the comparison of OFL and NOR in methanol, the interaction energies were similar in all sites (Fig. 3D). The comparable solubilities of OFL and NOR in methanol may be due to their similar interaction energies.

-40

O 22/18 N 5/11 F 14/17

B O 25/20 O 24/21 N 2/14

0

0

-10

-10

-20

-20

-30

-30

O 22/18 N 5/11 F 14/17

OFL-H2O OFL-CH3OH

-40

C

-40

NORNOR-CH3OH

D

Fig. 3. The comparison of interaction energies (kJ mol1) of hydrogen-bonded complexes for OFL with water/methanol (Panel A) and those of NOR with water/methanol (Panel B), Panel C is the comparison of interaction energies (kJ mol1) of OFL/NOR-water while Panel D is for comparison of interaction energies (kJ mol1) of OFL/NORmethanol. The detailed information for interaction energies were illustrated in Table S4.

327

H. Peng et al. / Chemosphere 103 (2014) 322–328

2

8

180

A

Cs -Eint

6

30 20

160

140

4 2

10 0 -0.2

180

B

0.0

0.2

0.4

0.6

0.8

1.0

0 -0.2

140 1.2

0.0

0.2

0.4

0.6

0.8

1.0

-Ein (KJ/mol)

Cs (×10 mg/L

40

100 1.2

fc Fig. 4. Relationship between interaction energies (Eint) of solute–solvent and the composition of water–methanol solvents. With the introduction of methanol, the water molecules formed hydrogen bonds with solutes were successively replaced by methanol, following the sequence of hydrogen bond lengths. The calculated interaction energies were plotted with fc. Panel A is for OFL and Panel B is for NOR.

2003). A more useful calculation is for a solute molecule surrounded by solvent molecules. The possible sites for hydrogen bond are the primary interaction sites. The calculation started as all the sites were occupied by water molecules, and the geometrical structures were optimized to obtain the interaction energy, Eint. With the introduction of methanol, the water molecules formed hydrogen bonds with solutes were successively replaced by methanol, following the sequence of the calculated Eint. The number ratios of solvent (water or methanol) associated solute molecules were calculated and plotted with fc (Fig. 4). It seems that the variation of solubilities with fc differed markedly from the variation of Eint. An increase of Eint was observed when the last two water molecules (at the sites of –N5 and –N2 for OFL, and –O18 and –N11 for NOR) were replaced by methanol for both solutes. More systematic experiments are needed to understand this phenomenon. But the inconsistence between the calculated Eint and solubility may be understood from the following reasons: (1) More solvent molecules may interact with solute besides the calculated ones (7 solvent molecules for NOR, and 6 solvent molecules for OFL). The physical vacancy around the solute molecules will be occupied by solvent molecules, and intra-molecular interactions (such as van der Waals interactions) are also expected. (2) When hydrogen bond was considered for a specific site, one molecule to one molecule interaction was assumed. However, in the real solution, one molecule may interact with several molecules through hydrogen bond, as for pure water system. In our and literature calculations, at least the above two conditions were not considered. 4. Conclusions The sorption OFL and NOR on CNTs from water–methanol cosolvent was comparable, showing decreased sorption with increasing methanol fraction. The log-linear cosolvency model was usually applied to compare the cosolvent–sorbent and solute–cosolvent interactions. However, the nonmonotonic variation of NOR solubility in water–methanol cosolvent failed the application of this loglinear cosolvency model. This work further showed the role of hydrogen bonds in OFL and NOR solubilization in the cosolvent. DFT calculation was performed at all the possible hydrogen bonds in OFL and NOR with water or methanol. The hydrogen bond length differed greatly for OFL and NOR at O22/18, N5/11, and O24/21 in water, but was rather similar in methanol. The calculated bond length and one solute molecule-one solvent molecule interaction energy provided useful information to understand the solubility change with methanol fractions as well as the difference between OFL and NOR. However, when the calculation was performed at the situation when all the possible hydrogen bond in OFL or NOR was occupied, the calculated interaction energy was not consistent with the measured solubility change with methanol volume

fraction. Similar sorption but different solubility of NOR and OFL from water–methanol cosolvent suggested that sorbate-solvent interaction does not control their sorption. Acknowledgments This research was supported by National Scientific Foundation of China (41173124, 41222025, and 41273138) Program for New Century Excellent Talents in University, Chinese Ministry of Education, Recruitment Program of Highly-Qualified Scholars in Yunnan (2010CI109), and a Research Fund for Senior Visiting Scholar at State Key Laboratory of Pollution Control and Resource Reuse (PCRRF11032). Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.chemosphere. 2013.12.025. References Abraham, M.H., Le, J., 1999. The correlation and prediction of the solubility of compounds in water using an amended solvation energy relationship. J. Pharm. Sci. 88, 868–880. Acelas, N.Y., Mejia, S.M., Mondragón, F., Flórez, E., 2012. Density functional theory characterization of phosphate and sulfate adsorption on Fe-(hydr)oxide: reactivity, pH effect, estimation of Gibbs free energies, and topological analysis of hydrogen bonds. Comput. Theor. Chem.. Becke, A.D., 1993. Density-functional thermochemistry. III. The role of exact exchange. J. Phys. Chem., 5648. Biserka, P., Kresimir, M., 2008. The nature of hydrogen bond: new insights into old theories. Acta Chim. Slov. 55, 692–708. Boonsaner, M., Hawker, D.W., 2010. Accumulation of oxytetracycline and norfloxacin from saline soil by soybeans. Sci. Total Environ. 408, 1731–1737. Bouchard, D.C., 2002. Cosolvent effects on sorption isotherm linearity. J. Contam. Hydrol. 56, 159–174. Chiou, C.T., Porter, P.E., Schmedding, D.W., 1983. Partition equilibriums of nonionic organic compounds between soil organic matter and water. Environ. Sci. Technol. 17, 227–231. Desai, K.G.H., Kulkarni, A.R., Aminabhavi, T.M., 2003. Solubility of rofecoxib in the presence of methanol, ethanol, and sodium lauryl sulfate at (298.15, 303.15, and 308.15) K . J. Chem. Eng. Data 48, 942–945. Efremenko, I., Sheintuch, M., 2006. Predicting solute adsorption on activated carbon: phenol. Langmuir 22, 3614–3621. Gandhi, P.J., Murthy, Z.V.P., 2012. Measurement of solubility of mitomycin C in ethanol–water solutions at different temperatures. Thermochim. Acta 545, 163–173. Goyne, K.W., Chorover, J., Kubicki, J.D., Zimmerman, A.R., Brantley, S.L., 2005. Sorption of the antibiotic ofloxacin to mesoporous and nonporous alumina and silica. J. Colloid Interface Sci. 283, 160–170. Hobza, P., šponer, J., Reschel, T., 1995. Density functional theory and molecular clusters. J. Comput. Chem. 16, 1315–1325. Hyung, H., Fortner, J.D., Hughes, J.B., Kim, J.-H., 2006. Natural organic matter stabilizes carbon nanotubes in the aqueous phase. Environ. Sci. Technol. 41, 179–184.

328

H. Peng et al. / Chemosphere 103 (2014) 322–328

Lam, C.-W., James, J.T., McCluskey, R., Arepalli, S., Hunter, R.L., 2006. A review of carbon nanotube toxicity and assessment of potential occupational and environmental health risks. Crit. Rev. Toxicol. 36, 189–217. Lee, C., Yang, W., Parr, R.G., 1988. Development of the Colle–Salvetti correlationenergy formula into a functional of the electron density. Phys. Rev. B 37, 785– 789. Lorphensri, O., Intravijit, J., Sabatini, D.A., Kibbey, T.C.G., Osathaphan, K., Saiwan, C., 2006. Sorption of acetaminophen, 17 alpha-ethynyl estradiol, nalidixic acid, and norfloxacin to silica, alumina, and a hydrophobic medium. Water Res. 40, 1481– 1491. Majerz, I., 2012. Directionality of inter- and intramolecular OHO hydrogen bonds: DFT study followed by AIM and NBO analysis. J. Phys. Chem. A 116, 7992–8000. Masciangioli, T., Zhang, W.-X., 2003. Environmental technologies at the nanoscale. Environ. Sci. Technol. 37, 102A–108A. Miyako, Y., Khalef, N., Matsuzaki, K., Pinal, R., 2010. Solubility enhancement of hydrophobic compounds by cosolvents: role of solute hydrophobicity on the solubilization effect. Int. J. Pharm. 393, 48–54. Nishio, M., Umezawa, Y., Honda, K., Tsuboyama, S., Suezawa, H., 2009. CH/p hydrogen bonds in organic and organometallic chemistry. CrystEngComm 11, 1757–1788. Pan, B., Sun, K., Xing, B., 2010. Adsorption kinetics of 17 alpha-ethinyl estradiol and bisphenol A on carbon nanomaterials. II. Concentration-dependence. J. Soils Sedim. 10, 845–854. Park, H.R., Kim, T.H., Bark, K.M., 2002. Physicochemical properties of quinolone antibiotics in various environments. Eur. J. Med. Chem. 37, 443–460. Peng, H., Pan, B., Wu, M., Liu, R., Zhang, D., Wu, D., Xing, B., 2012a. Adsorption of ofloxacin on carbon nanotubes: solubility, pH and cosolvent effects. J. Hazard. Mater. 211–212, 342–348. Peng, H., Pan, B., Wu, M., Liu, Y., Zhang, D., Xing, B., 2012b. Adsorption of ofloxacin and norfloxacin on carbon nanotubes: hydrophobicity- and structurecontrolled process. J. Hazard. Mater. 233, 89–96.

Picó, Y., Andreu, V., 2007. Fluoroquinolones in soil—risks and challenges. Anal. Bioanal. Chem. 387, 1287–1299. Schoone, K., Smets, J., Houben, L., Van Bael, M.K., Adamowicz, L., Maes, G., 1998. Matrix-isolation FT-IR studies and theoretical calculations of hydrogen-bonded complexes of molecules modeling adenine tautomers. 1. H-bonding of benzimidazoles with H2O in Ar matrices. J. Phys. Chem. A 102, 4863–4877. Taha, M., Lee, M.-J., 2010. Buffer interactions: solubilities and transfer free energies of TRIS, TAPS, TAPSO, and TABS from water to aqueous ethanol solutions. Fluid Phase Equilib. 289, 122–128. Tasis, D., Tagmatarchis, N., Bianco, A., Prato, M., 2006. Chemistry of carbon nanotubes. Chem. Rev. 106, 1105–1136. Ulu, S.T., 2009. Rapid and sensitive spectrofluorimetric determination of enrofloxacin, levofloxacin and ofloxacin with 2,3,5,6-tetrachloro-pbenzoquinone. Spectrochim. Acta Part A Mol. Biomol. Spectrosc. 72, 1038–1042. Van Bael, M.K., Smets, J., Schoone, K., Houben, L., McCarthy, W., Adamowicz, L., Nowak, M.J., Maes, G., 1997. Matrix-isolation FTIR studies and theoretical calculations of hydrogen-bonded complexes of imidazole. A comparison between experimental results and different calculation methods. J. Phys. Chem. A 101, 2397–2413. Wendler, K., Thar, J., Zahn, S., Kirchner, B., 2010. Estimating the hydrogen bond energy. J. Phys. Chem. A 114, 9529–9536. Zhang, D., Pan, B., Zhang, H., Ning, P., Xing, B., 2010. Contribution of different sulfamethoxazole species to their overall adsorption on functionalized carbon nanotubes. Environ. Sci. Technol. 44, 3806–3811. Zorita, S., Martensson, L., Mathiasson, L., 2009. Occurrence and removal of pharmaceuticals in a municipal sewage treatment system in the south of Sweden. Sci. Total Environ. 407, 2760–2770.