Copper removal from semiconductor CMP wastewater in the presence of nano-SiO 2 through biosorption

Copper-bearing wastewater from chemical mechanical planarization (CMP) is a typical semiconductor development byproduct. How to effectively treat Cu 2 þ in the CMP wastewater is a great concern in the microchip manufacturing industry. In this study, we investigated the potential for the microbial removal of Cu 2 þ by a multiple heavy metal-resistant bacterium Cupriavidus gilardii CR3. The environmental factors, including pH, nano-SiO 2 , ionic strengths, and initial concentrations of Cu 2 þ , and adsorption times on the bioremoval of Cu 2 þ in CMP wastewater were optimized. Under optimal condition, the maximum biosorption capacity for Cu 2 þ was 18.25 mg g (cid:2) 1 and the bioremoval rate was 95.2%. The Freundlich model is described well for the biosorption of Cu 2 þ in CMP wastewater in the presence of nano-SiO 2 ( R 2 ¼ 0.99). The biosorption process obeyed the pseudo-second-order kinetic equation ( R 2 > 0.99). In the column experiment, the advection – dispersion – retention model ﬁ tted the breakthrough curve of all experiments well ( R 2 > 0.95). The attachment coef ﬁ cient in the sand matrix coated by CR3 bio ﬁ lm was 2.24 – 2.80 times as that in clean sand. Overall, C. gilardii CR3 is a promising candidate to remove Cu 2 þ from CMP wastewater. Nano-SiO 2 in CMP wastewater did not inhibit the bioremoval of Cu 2 þ but showed a slight promotion effect instead. 2 þ . (cid:129) The Freundlich model is better for describing the biosorption of Cu 2 þ CMP wastewater in the presence of Nano-SiO 2 relative to the Langmuir model.


INTRODUCTION
Increasing demand for advanced semiconductor products or microchips contributes to economic development in nations around the globe; however, it is also associated with a series of environmental issues. A major environmental problem is the large volumes of wastewater containing complicated contaminant composites generated from semiconductor manufacturing. Chemical mechanical planarization/ polishing (CMP) is a vital process in semiconductor manufacturing for dielectrics and metal layer planarization.
In CMP, a considerable quantity of chemicals and ultrawater is needed to remove the residues of nanoparticles and chemicals on the semiconductor devices and achieve dielectrics and metal layer planarization. Thus, the resulting CMP wastewater contains various contaminants, which accounts for 40% of the entire semiconductor production process wastewater (Lai & Lin ). Once treated appropriately, CMP wastewater can be reclaimed and pressures on water resources can be relieved in the semiconductor industry.
In CMP wastewater, cupric ions (Cu 2þ ) pose a great concern due to their persistent nature, accumulation tendency, and biological toxicity. Cu 2þ in CMP wastewater originated from copper metallization, as well as the application of CuSO 4 as oxidizing agents during the CMP. The concentrations of Cu 2þ in CMP wastewater generally are presented as parts per million although the concentration of Cu 2þ fluctuates in different ranges, for example, 5-100 mg L À1 (Maketon ). The reported concentration of Cu 2þ in CMP wastewater is much higher than the level of copper toxicity toward aquatic organisms at 5-10 parts per billion (ppb) (Chua et al. ). Accordingly, appropriate treatment of Cu 2þ in CMP wastewater is necessary to meet the local discharge standards of pollutants for the semiconductor industry. In the districts of the developed semiconductor industry, the permissible range of Cu 2þ is from 0.2 to 1.0 mg L À1 depending on the receiving waters.
Cu 2þ in CMP wastewater is usually treated by the electrochemical method (Yang & Tsai ), ion exchange resin method (Maketon & Ogden ), and membrane filtration (Su et al. ). However, the high initial costs and additional chemicals hamper the application of physicalchemical methods. Alternatively, biosorption by growing cells can offer advantages for the removal of heavy metal ions at low concentrations (Malik ). In the case of Cu 2þ treatment in CMP wastewater, biosorption of Cu 2þ in CMP wastewater has been reported (Stanley & Ogden ; Mosier et al. ). Our previous study also investigated the characteristics of C. gilardii CR3 to treat Cu 2þ in CMP wastewater (Yang et al. ). However, the synthesized CMP wastewater used in the previous studies focused on the bioremoval of Cu 2þ . The CMP requires nanoparticles as abrasive particles, for example, nanoparticle SiO 2 (nano-SiO 2 ), which is the common nanoparticle in the CMP, usually ranging from 0.05 to 0.50 g L À1 (Yang & Tsai ). But the treatment of Cu 2þ rarely varies based on the presence of nano-SiO 2 in CMP wastewater. Hence, it should be considered whether the nano-SiO 2 can affect the biosorption of Cu 2þ when CMP wastewater is treated.
In this study, we investigated the bioremoval of Cu 2þ by a copper-resistant bacterium C. gilardii CR3 in the presence of nano-SiO 2 . In the batch adsorption system, the treatment conditions were optimized and the bioremoval mechanisms of Cu 2þ were examined. Further, the transport and retention of Cu 2þ was studied in a fixed-bed column packed with sand coated by strain CR3.

Bacterium and synthetic wastewater
The bacterium C. gilardii CR3 used in this study was isolated in our previous study (Yang et al. ). The cryopreserved C. gilardii CR3 cells were inoculated into a 100 mL Luria-Bertani (LB) medium and cultured in a constant temperature shaker at 28 C for 24 h. Then the resurrected bacterial cells in 1 mL were transferred into a new 100 mL LB media. The suspension was centrifuged at 10,000 r min À1 for 5 min to collect the living cells. The received cells by centrifugation were washed with phosphate-buffered saline three times and the resulting cell suspension was used for the batch and column experiments. The growth curve of the strain in the LB medium is shown in Supplementary Material, Figure S1.
To stimulate the real CMP wastewater containing Cu 2þ (Cu-CMP wastewater), the synthetic Cu-CMP wastewater was prepared with CuCl 2 ·H 2 O, nano-SiO 2 (50 nm), and deionized water.

Batch experiments
The effects of pH and nano-SiO 2 concentration, ionic strength, initial concentrations of Cu 2þ , and contact time on the biosorption of Cu 2þ by C. gilardii CR3 were studied.
The concentration of Cu 2þ used in batch experiments was 0.3 mM Cu 2þ except for the test of Cu 2þ initial concentrations. First, the pH (4.0, 5.0, and 6.0) and the concentration of nano-SiO 2 (0.1, 0.3, and 0.5) in CMP wastewater were optimized. Afterwards, the effects of ionic strength, initial concentrations of Cu 2þ , and contact time on the biosorption of Cu 2þ were conducted under the optimized conditions in the first step (pH ¼ 5, nano-SiO 2 ¼ 0.3 g L À1 ). The ionic strength was measured at 0, 1.0, 1.5, 2.0, 2.5, 3.0, 4.0, and 5.0 mM with a NaNO 3 solution. The initial Cu 2þ concentration was set to 10 levels, from 0.1 to 1 mM with intervals of 0.1. The contact times on Cu 2þ biosorption were conducted from 0 to 180 min at 28 C. In control experiments, the aqueous Cu 2þ was replaced with the same volume of deionized water. The removal rate and biosorption capacity of strain CR3 on Cu 2þ were calculated according to the following formula: where q e is the biosorption capacity of Cu 2þ by the unit of biosorbents in mg g À1 , C 0 and C e are the initial and equilibrium concentration, respectively, of Cu 2þ in the solution in mg L À1 , m is the mass of the biosorbents in g, and ν is the volume of Cu-CMP wastewater in the system in L. In this study, Langmuir and Freundlich models were used to describe the equilibrium process of biosorption in batch experiments. The models are expressed as follows: in which q e is represented as the biosorption capacity of Cu 2þ by the unit of biosorbent in mg g À1 , C e is the equilibrium concentration of Cu 2þ in the solution in mg L À1 , q m is the theoretical monolayer biosorption capacity in mg g À1 , n and K F are the Freundlich equilibrium constants, and b is the Langmuir equilibrium constant.
To study the biosorption of the system, the experimental results are fitted with pseudo-first-order kinetics and pseudosecond-order kinetics. The firstand second-order equations are expressed as follows: log(q e À q t ) ¼ log q e À k 1 2:303 t (5) where q t and q e , respectively, represent the biosorption time t and the amount of Cu 2þ adsorbed by the biosorbent when they reach biosorption equilibrium in mg g À1 and k 1 and k 2 are the first and second rate equation constants.
Scanning electron microscope, Fourier-transform infrared spectroscopy, and zeta potential A scanning electron microscope (SEM) (XL-30 ESEM FEG) was used to investigate the quartz sand, nano-SiO 2 , and C. gilardii CR3 with Extracellular Polymeric Substances (EPS) in this study.
Quartz sand was ultrasonically cleaned three times and incubated with strain CR3 for 24 h. The quartz sand, strain CR3, and sand-strain CR3 were prepared into a 1.0 g L À1 suspension. The Zeta potential of strain CR3, quartz sand, and sand-strain CR3 was measured before and after the biosorption of Cu 2þ using a Zeta potentiometer (Malvern Zetasizer Nano ZS90) under the pH 5.0, 0.3 mM Cu 2þ , and 0.3 g L À1 nano-SiO 2 condition.
The sand-strain CR3 biosorbent samples before and after biosorption of Cu 2þ were freeze-dried and ground with KBr at a ratio of 1:100. The sand-strain CR3 biosorbent samples were analyzed using Fourier-transform infrared (FTIR) spectroscopy (SHIMADZU FTIR-8400S).

Transport column experiments
The Cu 2þ transport experiments were conducted using an acrylic column under saturated flow conditions (Supplementary Material, Figure S2). The column was 2.2 cm in inner diameter and 11 cm in effective length. Quartz sand with a diameter of 0.7-1.0 mm was packed in the column as a porous medium, and the packing density was 1.39 g cm À3 .
Screens were installed at both ends of the column to prevent the packing from flowing out and distributing the flow Compact IC pro, Switzerland). Normalized effluent concentrations (C/C 0 ) were plotted against the number of PVs to obtain the breakthrough curves (BTCs).

Mathematical modeling for column experiments
The transport of Cu 2þ within the saturated sand column was described by an advection-dispersion-retention (ADR) model considering dynamic blocking and depth-dependent straining. The governing equations can be written as: where n is the porosity; C is the concentration of effluent Cu 2þ ; ρ is the bulk density of the porous media; S is the solid phase concentration adsorbed on the quartz sand; q is the flow rate; k att and k det are the first-order attachment coefficient and detachment coefficient, respectively; and ψ is a dimensionless function to account for the combined process of Langmuirian dynamics blocking and depthdependent straining (Bradford & Bettahar ). S max is the maximum solid phase particle concentration. The trace (Br À ) experiments were conducted before the Cu 2þ penetration experiments to estimate the hydraulic dispersion coefficient (D). The above ADR model was fitted with the penetration curve to obtain k att , k det , and S max parameters.
The Cu 2þ mass of water was calculated by the integral penetration curve BTCs based on the known Cu 2þ mass input, and then the total adsorption mass of Cu 2þ in the sand column was calculated according to the mass balance.

Measurement of Cu 2þ
The Cu 2þ concentration was measured by flame atomic absorption spectrometry (Varian 220FS, USA). All experiments were conducted in triplicate in this study. The data are given as mean values and the 95% confidence interval around the average value is indicated by error bars.

RESULTS AND DISCUSSION
Batch experimentseffects of pH, nano-SiO 2 , ionic strength, initial Cu 2þ concentration, and contact time The pH value of the aqueous solution is an essential parameter that could affect the metallic ions biosorption (Alavi et al.   Figure S4).   Table 2.

Biosorption-isotherms and kinetics
The fitting results of quasi-first-order kinetics and pseudo-second-order kinetics model are shown in Table 2.
The fitting result of the quasi-first-order kinetic model was poor (R 2 ¼ 0.53), thus the resulting q e value (0.271 mg g À1 ) is lower than the value of equilibrium biosorption capacity (18.88 mg g À1 ). The pseudo-second-order kinetic model obeyed the experimental data (R 2 > 0.99). Meanwhile, the q e value obtained by the pseudo-second-order kinetic is 18.97 mg g À1 , which is in good agreement with the equilibrium biosorption capacity.

Analysis of SEM, zeta potential, and FTIR
The SEM images of C. gilardii CR3 before and post Cu 2þ binding are shown in Figure 2. Compared to the sample without Cu 2þ (Figure 2 The Zeta potential of strain CR3, quartz sand, and the composite of quartz sand-strain CR3 were measured before and after the biosorption of Cu 2þ (Table 3). Among the three sorbents, the potential value of composite of quartz sand-strain CR3 was the most negative (À55.4 mV), followed by strain CR3 (À22.5 mV) and quartz sand (À45.9 mV) before biosorption of Cu 2þ , which indicated that composite biosorbent carried more negative charges than that of strain CR3 and quartz sand. Likewise, the potential value showed a similar trend with the before biosorption of Cu 2þ . Zeta potential represents the surface charge of the biosorbent and reflects the electrochemical properties of the biosorbent surface (van der Mei & Busscher ).
Thus, the results show that the composite of quartz sandstrain CR3 is favorable to bind more Cu 2þ as compared to strain CR3 and quartz sand because the composite possesses a stronger biosorption capacity for Cu 2þ .
The results of the FTIR analysis showed that the peak shape of the composite of quartz sand-strain CR3 remained unchanged before and after the biosorption of Cu 2þ (Figure 3).
In other words, Cu 2þ did not destroy the structure of the composite. In the composite of quartz sand-strain CR3, the major sorption peaks were as follows: 3,491 cm À1 corresponding to -NH and -OH stretching; 2,921 cm À1 corresponding to -CH 2 symmetrical or asymmetrical stretching; 1,082 cm À1 corresponding to C-OH/C-O-C/C-C/P ¼ O; and the remaining absorption peak corresponds to the SiO 2 peak position in quartz sand. The specific functional groups on the surface of  The hydrodynamic dispersion coefficient D was estimated as 0.28 ± 0.07 cm 2 min À1 (R 2 > 0.96). This fitted value of D was then used to model the transport of Cu 2þ , and the best-fit transport parameters in the model are listed in As shown in Figure 4, the coexisting nano-SiO 2 could facilitate the transport of Cu 2þ , since the retention ratio of Cu 2þ was 29.73% in clean sand without nano-SiO 2 and it decreased to 23.90% when nano-SiO 2 was introduced.
Nano-SiO 2 worked as the Cu 2þ -carrier and promoted Cu 2þ transport, which was further validated in the strain gilardii CR3 biofilm, the k att increased 2.24-2.80 times the k att in clean sand. The k det was nearly 3-5 orders of magnitude lower than k att , which confirmed that the remobilization of previously retained Cu 2þ could be inconsiderable and the slight differences of k det under these scenarios could be ignored. Overall, the presence of nano-SiO 2 could enhance Cu 2þ transport and thus may increase spreading risk. By contrast, C. gilardii CR3 biofilm in the sand matrix can be a barrier to Cu 2þ breakthrough.
The aforementioned facilitated bioremoval of Cu 2þ by strain CR3 was confirmed by additional continuous flow column experiments, which was described in Section S1 in Supplementary Material. We employed two kinds of CMP  wastewaters to be injected into the columns continuously: one only contained Cu 2þ and one contained Cu 2þ and nano-SiO 2 together. The results showed that the relative effluent concentration of Cu 2þ (C/C 0 ) was higher in quartz sand without strain CR3 compared with that in CR3coated sand matrix (Supplementary Material, Figure S6) for both of the two CMP wastewater, while the average C/ C 0 at t ¼ 60 min was 0.90 for the former and 0.81 for the latter. Note that the biosorption in these breakthrough columns was weaker than those in batch adsorption system; it can be attributed to the shorter hydraulic retention time in the columns (less than 12 min). Nevertheless, it still could be seen that the Cu 2þ transportation was retarded by CR3 biofilm. By contrast, the differences between different C/C 0 curves in the same medium with/o nano-SiO 2 were less than 2.20% (Supplementary Material, Figure S6).  and Cu(OH) þ , which means lower bioremoval of Cu 2þ . If the pH value is over 5.0, some Cu 2þ will present as Cu (OH) 2 according to the solubility product constant (2.2 × 10 À20 ). Since the aim of this study is to estimate the capability of bacterium to remove Cu 2þ , higher pH condition (>5.0) is not considered although the insolubility of copper hydroxide can lead to small extra Cu 2þ removal. When pH value ranges between 2.2 and 5.0, Cu 2þ and Cu (OH) þ are more preferable in the aqueous solution. As a result, these copper species will interact with the function groups (Figure 3), for example -SH, -OH, and -COOH, on the surface of microbial cells and increasing the binding of Cu 2þ . In addition to the surface sorption, bioaccumulation and biotransformation also contribute to some extent to the removal of Cu 2þ . Once the binding on the cell sites are fully occupied, bioaccumulation will contribute to the bioremoval of Cu 2þ (Ayangbenro & Babalola ). At the same time, the intracellular Cu 2þ might be effluxed out to keep the level of intracellular Cu 2þ below the toxicity threshold (Machalová et al. ). It should be noted that the synthesized CMP wastewater used in the above-mentioned studies contained Cu 2þ without nano-SiO 2 . According to our knowledge, the finding in our study is the first to indicate that coexisting nano-SiO 2 does not inhibit the biosorption Cu 2þ by strain CR3.

CONCLUSIONS
In this study, bioremoval of Cu 2þ from CMP wastewater in the presence of nano-SiO 2 was investigated through batch adsorption experiments and transport column experiments using a typical copper-resistant bacterium C. gilardii CR3.
The Freundlich isotherm is suitable for describing the biosorption process in Cu 2þ -nano-SiO 2 -CR3 ternary system.
The biosorption kinetic data fit well with the pseudosecond-order equation. Overall, the maximum biosorption capacity for Cu 2þ was 18.25 mg g À1 and the bioremoval rate was as high as 95.2%, displaying high removal efficacy of strain CR3. In the transport column experiments, the BTCs fit well with the ADR model. The attachment coefficient in the sand matrix coated by CR3 biofilm was 2.24-2.80 times as that in clean sand, confirming that CR3 biofilm can be a barrier to retard Cu 2þ transport. In particular, in all experiments, the presence of nano-SiO 2 did not inhibit the bioremoval of Cu 2þ from CMP wastewater but showed a slight promotion effect instead. Therefore, C. gilardii CR3 is a promising microbe biosorbent to remove Cu 2þ from CMP wastewater even in the presence of nano-SiO 2 .
ACKNOWLEDGEMENT This work was funded by the National Natural Science Foundation of China (no. 51678122, no. 52070037, and no. 51978135).