Core–shell TiO2@C ultralong nanotubes with enhanced adsorption of antibiotics

As materials capable of adsorbing antibiotics continue to be developed, composite adsorbents have been shown to offer advantages over mono-material adsorbents. In this work, ultralong titanium dioxide@carbon nanotubes were prepared by a simple hydrothermal treatment, followed by carbonization. The composite material is able to adsorb three different categories of antibiotics, including tetracycline (TC), ofloxacin (OFO) and norfloxacin (NFO). The adsorption results show that the adsorption properties of composite materials have been greatly improved compared with single inorganic adsorbent materials, for which the adsorption capacities are 240 mg g 1 (TC), 232 mg g 1 (OFO), and 190 mg g 1 (NFO), respectively. The adsorption mechanism is consistent with a Langmuir pseudo-first-order kinetic model.


Introduction
With the discovery of antibiotics, such as penicillin, for human beings to ght infectious diseases, there is a common phenomenon of abusing antibiotics to cope with various illnesses. 1 However, antibiotics are difficult to metabolize in a short period of time in animals, plants and humans. Simultaneously, emissions from pharmaceutical companies have also caused enormous environmental pollution. 2,3 Thus, a large amount of used antibiotics went to the soils, rivers, animals, etc. The accumulation of these residual antibiotics not only pollutes the environment, but also causes harm to human health. 4 With the emergence of super bacteria, there is serious concern that society will lack the required antibiotics in future. 5 The warnings of scientists have attracted the attention of the public and governments. Therefore, searching for effective and non-toxic methods to remove antibiotic pollutants has become a hot research topic.
There are a number of methods which aim to decrease the amount of antibiotics in the environment, including adsorption, 6 photocatalysis, 7 electrochemical oxidation, 8 advanced oxidation 9 and biodegradation. 10 Among these methods, adsorption is a simple and cheap way to solve the problem of antibiotic pollution and can be divided into two categories: chemisorption and physisorption. Chemisorption requires high activation energy and usually higher temperatures, while physisorption is usually caused by molecular force and can occur at lower temperature. Furthermore, the chemisorption reaction may be undetectable at low temperature or pressure. 11 As characteristic of physisorption, for gases at temperatures above their critical temperature, adsorption is conned to a monolayer (unimolecular). 12 Therefore, the adsorption capacity is related to the specic surface area and structures of the adsorbents. The effects of the specic surface area and structure of different types of single adsorbent materials and composite adsorbent materials on the adsorption capacity have been studied in depth. 13 Promising materials, such as zeolites for adsorption of tetracycline, 14 attapulgite/carbon composites for adsorption of tetracycline, 15 and oil shale powders 16 and functional MOFs for adsorption of noroxacin, 17 were reported. A large number of experimental results show that the specic surface area and the structure of composites are the key factors that make adsorption more effective than with a single material. Therefore, there is an increasing focus on developing composite adsorbent materials with an optimized structure and high specic surface area via simple preparation processes amenable to mass production.
In this work, ultralong TiO 2 nanotubes were prepared using a simple hydrothermal treatment. In order to sustain the ultralong structure and increase the specic surface area, the ultralong TiO 2 nanotubes were covered with an ultrathin carbon layer by carbonizing ethanol, and thus a core-shell TiO 2 @C ultralong nanotube composite was formed. Three common antibiotics, i.e., tetracycline (TC), ooxacin (OFO) and nor-oxacin (NFO), were adsorbed by the TiO 2 @C composite, and the adsorption capacities were measured to be 240 mg g À1 , 232 mg g À1 , 190 mg g À1 , respectively, which are higher than those of many of the reported single and composite adsorbent materials. Therefore, the core-shell TiO 2 @C nanotubes are promising adsorbent materials for antibiotic pollutants. This study sheds light on the mechanism of metal oxide/carbon hybrids for adsorbing antibiotics and identies routes to improve performance.

Experimental section
Preparation of ultralong titanium dioxide nanotubes Firstly, 0.5 g P 25 (a commercial titanium dioxide which is an anatase and rutile two-phase mixture nanopowder with an average particle diameter of 25 nm) was added into 40 mL of 10 mol L À1 NaOH solution and dispersed in a 50 mL reaction kettle, which was ultrasonicated for 5 min to obtain a homogeneous solution. The reaction kettle was sealed and put into an oil bath at 120 C for 48 h with magnetic stirring (Fig. 1a). When the reaction nished, the reactor was taken out and cooled to room temperature. The obtained white products were washed with 0.1 mol L À1 HNO 3 and centrifuged several times until the pH reached 7. The nal products were washed with deionized water and placed in a vacuum drying oven at 60 C for 8 h.

Preparation of TiO 2 @C core-shell structures
The as-prepared products were placed in the middle of a tube furnace, N 2 gas was purged through anhydrous ethanol and then into the tube furnace, and the reaction temperature was set at 400 C for 10 h under a nitrogen atmosphere. Black products (TiO 2 @C composite) were obtained as the reaction nished. The experimental schematic diagram is shown in Fig. 1b.

Characterization
The prepared TiO 2 @C composites were characterized by scanning electron microscopy (SEM, S-4800, Japan) and transmission electron microscopy (TEM, JEM-2100F, Japan). The crystal structure and components of the material were determined using an X-ray diffractometer (XRD, Rigaku, Cu-Ka radiation, Japan). The specic surface area and the pore size distribution of the material were determined from the Brunauer-Emmett-Teller (BET) test (ASAP 2020, Micromeritics, America). The surface chemical structure was determined by Fourier transform infrared spectroscopy (FT-IR) (6700, Nicolet, USA) and Raman spectroscopy (Lab RAM, HR800) with 532 nm laser light. The absorbance of the TiO 2 @C composite materials was measured using a UV spectrophotometer (G-9, Rangqi, China). The functional groups and the elemental states on the surface of the adsorbents were examined by X-ray photoelectron spectroscopy (XPS) analysis, which was conducted on a Kratos Axis Ultra DLD spectrometer, using monochromated Al Ka Xrays at a base pressure of 1 Â 10 À9 Torr.

Results and discussion
The result of XRD analysis is shown in Fig. 2; the main diffraction peaks could be indexed to the (101), (103), (200), (105), (213), (116) and (107) reections, corresponding to anatase phase TiO 2 with the JCPDS card no. 01-0562. The result also conrms that C exists in the TiO 2 products (marked with red asterisks). The C species were further tested by Raman spectroscopy and the result is shown in Fig. S1. † The peak intensity ratio between the G band and D band (I G /I D ) is 1.1796, indicating that the proportion of carbon defect structures in the carbon structure is relatively small and the dominant component in the carbon structures is the sp 2 .
The low-and high-magnication SEM images of the TiO 2 @C products are shown in Fig. 3a and b. From the SEM images, it   can be seen that the length of the TiO 2 @C nanotubes can be up to several micrometers, representing an 'ultralong' structure. From the TEM image of Fig. 3c, it can be conrmed that the composite material is a tubular nanostructure. More detailed information on the material can be obtained from Fig. 3d; the outer diameter of the nanotubes is $35 nm, and the inner diameter is $15 nm, and the thickness of the carbon layer is 5-6 nm (Fig. 3e). This shows that the carbon layer was successfully coated on the outside of the TiO 2 nanotubes, and the interplanar spacing of 0.35 nm indicates the (101) lattice plane of the TiO 2 crystal.
The detailed specic surface area of the TiO 2 @C core-shell nanotubes can be obtained from the N 2 adsorption-desorption curve (Fig. 4a). The specic surface area of TiO 2 @C can be calculated from the N 2 adsorption-desorption isotherm, and the BET value is 346.26 m 2 g À1 , which is higher than that of the single TiO 2 (Fig. S2 in the ESI †). Fig. 4b shows the pore size distribution of the TiO 2 @C nanotubes, with a main diameter of around 15 nm, which corresponds to the diameter of the nanotubes. The diameter of $60 nm can be attributed to the gaps between the nanotubes.
The adsorption of the three kinds of antibiotics is based on the classical monolayer Langmuir model and the multilayer Freundlich adsorption model. The Langmuir equation depicts the process of randomly adsorbing species onto the surface of the adsorbent. The linearization equation of the Langmuir adsorption model is shown as follows: 18 where C e is the balance concentration (mg L À1 ), K L is the adsorption constant (L mg À1 ), q m is the maximum adsorption capacity (mg g À1 ), and q e represents the quantity (mg g À1 ) of antibiotics adsorbed at equilibrium. The Freundlich adsorption model is an empirical equation that is simple and widely used. It can simulate a multi-layer adsorption process, which means that the adsorption sites on the adsorption surface are not uniform. The linearized form of the Freundlich adsorption model equation is as follows: 19 where K F and n represent the Freundlich constant [(mg g À1 ) (L mg À1 ) 1/n ] and adsorption intensity of the adsorbents, respectively. [20][21][22] The mass of the TiO 2 @C materials used in this study was 5 mg, for adsorption of 5-40 mg L À1 tetracycline (TC), 5-40 mL ooxacin (OFO) and 2-20 mg L À1 noroxacin (NFO). Adsorption equilibrium was achieved aer magnetic stirring at room temperature. Fig. 4 shows the experimental data of the TiO 2 @C nanotubes adsorbing the three kinds of antibiotics: OFO, TC and NFO. Fig. 5a and b are the Langmuir model tting charts, and Fig. 5c and d show the corresponding Freundlich model tting charts. The tting data are shown in Table 1. From the correlation coefficient (R 2 ) value of the two models, it can be concluded that the three kinds of antibiotics adsorbed by the material are more consistent with the Langmuir model. The maximum adsorption capacities of OFO, TC and NFO, calculated using the Langmuir model, are 232 mg g À1 , 240 mg g À1 , and 190 mg g À1 , respectively. The experimental results show that the adsorption amount of this material is greatly improved compared with other inorganic materials, as shown in Table 2.

Adsorption kinetic model
The adsorption kinetics of the three kinds of antibiotics were studied. The pseudo-rst-order model was based on the assumption that the adsorption was controlled by diffusion. 29 The pseudo-rst-order dynamic model is as follows: 30 ln(q e À q t ) ¼ ln q e À K 1 t (3) Fig. 4 (a) The N 2 adsorption-desorption isotherm and (b) pore diameter distribution of the TiO 2 @C nanotubes.  The pseudo-second-order kinetic model assumes that the adsorption rate is determined by the square value of the number of adsorbed vacancies on the surface of the adsorbent. The pseudo-second-order dynamic model is as follows: 31 where q e is the concentration of adsorption at equilibrium (mg g À1 ), q t is the amount of adsorption (mg g À1 ) at time t, K 1 is the pseudo-rst-order adsorption rate constant, and K 2 is the pseudo-second-order adsorption rate constant. 17 The tting data of the dynamics of the three kinds of antibiotics are given in Table 3. Fig. 6 shows the pseudo-rst-order and pseudo-second-order kinetic models of the three antibiotics, and the intra-particlediffusion model. As Table 3 shows, the pseudo-rst-order model best describes the adsorption mechanism. Thus, the adsorption process of TiO 2 @C adsorbing OFO, TC and NFO is mainly controlled by diffusion. 32 In order to obtain more information on the adsorption process, data are tted to the intra-particle-diffusion model, as follows: 33 Adsorption can be divided into two processes: adsorbent surface adsorption and slow pore diffusion. The particle diffusion model is the most suitable for describing the dynamics of particles in the particle diffusion process. 34 The linear tting of q t and t 1/2 gives a straight line that passes through the origin and shows that particle diffusion is the rate determining step. 35 From Fig. 6c, it can be seen that the TiO 2 @C particle diffusion model is divided into three stages; the rst phase is the diffusion of the boundary layer, which is mainly the external mass transfer of the adsorbent. The second phase refers to the internal diffusion of particles, which is mainly the transport of microstructures of adsorbents, such as holes. These two stages make up the intra-particle diffusion model. Because the contact angle between the material itself and the solution is different, the time needed for different materials to reach adsorption equilibrium is different. The third phase is usually not considered as a rate-controlling step, generally referring to the procession of adsorption being close to adsorption equilibrium, which means that the active sites of the internal and external surfaces are mostly occupied. Compared with single titanium dioxide nanotube materials, the contact angle of the composites changed due to the synergistic effect of titanium dioxide nanotubes and the carbon layer, which greatly improved the adsorption performance of the material 36 (Fig. S3 †).
In order to understand the change that has occurred on the surface of the TiO 2 nanotubes, FTIR and XPS tests were carried out and TC was selected as an example. Fig. 7a shows the FTIR comparison, before and aer the adsorption of TC. The  absorption peak at 2920 cm À1 is attributed to the C-H bond. The hydrogen ion was introduced into the soaking TiO 2 nanotubes from dilute nitric acid, combined with a small amount of carbon into the C-H bond. The absorption peak at 1384 cm À1 is the result of the N-O bond introduced by TiO 2 nanotubes with dilute nitric acid. 5 The smaller absorption peaks at 1066 cm À1 and 903 cm À1 are due to the telescopic vibration of the C-O bond. The absorption peak of TiO 2 @C at 1609 cm À1 is the characteristic peak of the amide group. Aer adsorption of TC, the peak shied to 1630 cm À1 , indicating that TC was successfully adsorbed on TiO 2 @C. The sharp peak at 468 cm À1 is the result of the vibration of the Ti-O bond of titanium dioxide nanotubes. In addition, the carbonyl group is formed by the combination of some oxygen elements, which makes the Ti-O bond peak of the TiO 2 @C composite blunt and the peak shis to 497 cm À1 . The analysis of FT-IR results showed that the adsorption of TiO 2 @C on TC was mainly physical adsorption, which was in accordance with the pseudo-rst-order kinetic model. To further prove that TC has been adsorbed by TiO 2 @C, XPS spectra (Fig. S4 †) of the C 1s peaks of adsorbents before and aer adsorption are compared. Aer deconvolution, the C 1s signal can be separated into a C]C/C-C peak at 284.9 eV, a C-O peak at 292.3 eV and a C]O peak at 292.9 eV (Fig. 8b). It is noted that the content of C]C/C-C is much lower than that of C]O. However, aer the TC antibiotic was adsorbed by TiO 2 @C, the adsorbed C 1s signal is deconvoluted to a C]C/C-C peak at 284.5 eV, a C-O peak at 285 eV and a C]O peak at 286.2 eV. Simultaneously, the content of C-C/C]C increased sharply. In contrast, the content of C]O is the lowest among the three kinds of chemical bonds. In summary, the increase of C]C/C-C is due to the benzene ring of TC. When the TC antibiotic was adsorbed by TiO 2 @C, the C]O would be replaced by C]C/C-C. Thus, it is noted from the XPS spectra that the TC antibiotic can be adsorbed by TiO 2 @C effectively. The D-R isotherm model can be applied to identify the type of adsorption, i.e., physical or chemical adsorption process. The linear D-R isotherm 37 models are as follow and the results are shown in Fig. 8: where q e and q m (mmol g À1 ) are the equilibrium antibiotics solution concentration on the adsorbent and maximum adsorption capacity, respectively, B D is a constant related to the adsorption energy (mol 2 J À2 ), 3 is the Polanyi potential, C e is the equilibrium concentration (mol L À1 ), R is the gas constant and T is the absolute temperature (K). The adsorption free energy (E) can be calculated according to Polanyi potential from the following equation: Compared with the value of R 2 , it is obvious that the Langmuir model is more t for the experimental data. As for the D-R equations, if the value of E is less than 8 kJ mol À1 , physical sorption is considered as the main mechanism, while chemical sorption will be expected if the value of E is in the range of 8-16 kJ mol À1 . As shown in Table 4, the three values of E are all less than 8 kJ mol À1 . This proves that physisorption is the main sorption mechanism. The results of this analysis yielded consistent pseudo-rst-order adsorption kinetics.
The thermodynamics parameters regarding antibiotic adsorption onto TiO@C, such as Gibbs free energy (DG), and the solid and liquid phases at equilibrium (K d ), can be calculated using the following equations: 38 Generally, if À20 < DG < 0 kJ mol À1 , physical adsorption is dominant and it changes to more negative values than Fig. 7 (a) The FTIR spectra of TiO 2 @C before and after adsorbing TC, (b) high-resolution XPS spectra of the C 1s peaks before and after adsorption of the TC antibiotic by TiO 2 @C.   À40 kJ mol À1 for chemical adsorption. As shown in Table 5, the calculated result is consistent with D-R equations and pseudo-rst-order adsorption kinetics.

Conclusions
Core-shell TiO 2 @C ultralong nanotubes were prepared using a solvothermal process followed by anhydrous ethanol carbonization. The TiO 2 @C exhibits a mesoporous structure and large specic surface area. The adsorption experiment proves that the TiO 2 @C composite adsorbent can adsorb three kinds of typical antibiotics (TC, OFO, and NFO) effectively. Moreover, the main determinant of the adsorption process is proved to be physical adsorption rather than chemical adsorption, and the adsorption process ts the Langmuir and pseudo-rst-order models. Therefore, the work provides a practical case for the future development of more composite adsorbents and lays a foundation for the industrialization of composite materials.

Conflicts of interest
There are no conicts to declare.