In Situ Synthesis of Carbon Nanotube–Steel Slag Composite for Pb(II) and Cu(II) Removal from Aqueous Solution

Methods and materials that effectively remove heavy metals, such as lead and copper, from wastewater are urgently needed. In this study, steel slag, a low-cost byproduct of steel manufacturing, was utilized as a substrate material for carbon nanotube (CNT) growth by chemical vapor deposition (CVD) to produce a new kind of efficient and low-cost absorbent without any pretreatment. The synthesis parameters of the developed CNT–steel slag composite (SS@CNTs) were optimized, and its adsorption capacities for Pb(II) and Cu(II) were evaluated. The results showed that the optimal growth time, synthesis temperature and acetylene flow rate were 45 min, 600 °C and 200 sccm (standard cubic centimeter per minute), respectively. The SS@CNTs composite had a high adsorption capacity with a maximum removal amount of 427.26 mg·g−1 for Pb(II) and 132.79 mg·g−1 for Cu(II). The adsorption proceeded rapidly during the first 15 min of adsorption and reached equilibrium at approximately 90 min. The adsorption processes were in accordance with the isotherms of the Langmuir model and the pseudo-second-order model, while the adsorption thermodynamics results indicated that the removal for both metals was an endothermic and spontaneous process. This study showed that compared with other adsorbent materials, the SS@CNTs composite is an efficient and low-cost adsorbent for heavy metals such as lead and copper.


Introduction
Heavy metals, such as lead (Pb), copper (Cu), cadmium (Cd), chromium (Cr) and mercury (Hg), are nonbiodegradable and highly toxic, and can accumulate in the biosphere via the food chain [1]. The accumulation of heavy metals in the human body can lead to cancer and other tissue lesions [2,3]. Compared with other heavy metals, lead and copper are not only widely used in industrial manufacturing but are also constantly exposed to people in daily life. Notably, lead and its compounds are defined as class II carcinogens by the International Agency for Research on Cancer (IARC), as they are a possible cause of lung and stomach cancers. Moreover, excessive intake of lead and copper is likely to cause serious damage to the kidney and nervous and respiratory systems [4][5][6]. Consequently, methods and materials that effectively remove lead and copper from wastewater are urgently needed.
At present, various methods are applied in the elimination of lead and copper from wastewater such as chemical precipitation, electrodialysis and membrane filtration [7][8][9][10][11]. Although these methods are effective in removing copper and lead from wastewater, they have their own limitations. The chemical precipitation method requires an additional operational cost in the removal of the hazardous residues produced by itself. The main drawback with electrodialysis is high energy consumption, while the high cost in production and the complexity of residue treatment limit the application of the membrane filtration method [12,13]. Compared to these methods mentioned above, adsorption has proved to be a better and more widely used approach because of its operational ease, high sorption capacity and safety for the environment [14,15]. A variety of absorbent materials have been used in the removal of lead and copper, including chitosan, zeolites, diatomite, activated carbon, carbonized sawdust and lignite. However, some of these adsorbents have low removal efficiency for lead and copper [13,[16][17][18]. Carbon nanotubes (CNTs), a tubular carbon-based nanomaterial, have been validated as a novel and effective adsorbent for lead and copper removal owing to their large specific surface area, abundant porosity, outstanding chemical stability and the potential to be modified by functional groups [15,18,19].
Unfortunately, their hydrophobicity and high aspect ratio makes CNTs more prone to agglomeration, which leads to insufficient contact with the metal ions. Combined with their high cost, the above weaknesses have limited the application of CNTs [20]. Most of the recent research has focused on the pretreatment of CNTs to improve their dispersion and adsorption performance including surface modification and physical dispersion methods. Rodriguez et al. utilized multiwalled carbon nanotubes (MWCNTs) oxidized by nitric acid and sulfuric acid to remove Cu(II), Mn(II) and Zn(II) from water and discovered that the dispersion and adsorption efficiency of MWCNTs were augmented [21]. Oliveira et al. applied a surface active agent and ultrasonic method for the dispersion of MWCNTs, which subsequently enabled a higher removal amount of Pb(II), Cu(II), Ni(II) and Zn(II) [22].
Nevertheless, these pretreatment processes of CNTs are complex and costly, preventing their large-scale application. Fortunately, an approach whereby CNTs are synthesized on other larger adsorbent particles to form a composite by the chemical vapor deposition method (CVD) appears to reduce the agglomeration of CNTs by way of being dispersed with the dispersion of the particles in water. Moreover, this approach also provides the feasibility of large-scale adsorption application owing to the characteristics of the CVD method. As reported by other researchers, the composite seems to possess a higher adsorption capacity owing to the combination of the advantages of CNTs and other absorbents. Wang et al. synthesized CNTs on the diatomite to form a composite adsorbent for the elimination of phenol from water, and the results showed that the uptake amount of phenol by the composite significantly increased [23]. However, these materials used as the substrate for the CNTs' growth need to be pretreated by being coated with a catalyst such as ferric nitrate, nickel nitrate or cobalt nitrate. This necessity complicates the synthesis process of the composite adsorbent, and easily causes secondary pollution due to the residual metal catalyst.
Steel slag, a low-cost byproduct of the steel industry, contains some iron-oxides (generally more than 30%, Table 1), which are regarded as an appropriate catalyst for the CNTs' growth. Hence, steel slag can be utilized as a substrate with no need for any other pretreatment. Moreover, as reported in other studies, steel slag also has a good removal effect for Pb(II) and Cu(II). Yang et al. researched the sorption capacity of steel slag for Pb(II), Cu(II) and Cd(II) in acid solution and simultaneously evaluated the compounds formed after adsorption. The results showed that steel slag provided a remarkable removal of metal ions by ion exchange, chemical precipitation and co-precipitation to form stable compounds [24]. Shi et al. reported that the physical absorption also made a contribution to the adsorption efficiency of steel slag that was determined by the pore structure and the surface area of steel slag particles [25]. Due to the fact of its direct assemblability with CNTs and its effective adsorbability for heavy metals, steel slag can be used as a substrate for CNTs' growth via CVD to form a new composite adsorbent. Moreover, considering waste reuse, steel slag is still recognized as a solid waste and has not been utilized well. Hence, further utilization of steel slag is an effective way to turn waste into treasure and protect the environment [26]. Therefore, this new kind of adsorbent, formed by synthesizing CNTs on steel slag, reveals a potential prospect in the application for the removal of Pb(II) and Cu(II). In this study, CNTs-steel slag composite (SS@CNTs) was directly synthesized for the first time by means of CVD without any other pretreatment and used as a new kind of efficient and low-cost adsorbent. First, the effects of synthesis conditions, including the growth time, synthesis temperature and acetylene flow rate on the yield; crystallinity; microstructure of CNTs as well as the specific surface area and porosity characters of the composite were studied. Second, the efficiency of the adsorption of Pb(II) and Cu(II) by the synthesized SS@CNTs was evaluated and compared with other adsorbents. The adsorption process (i.e., adsorption kinetics, adsorption isotherms and adsorption thermodynamics) was investigated. Finally, the adsorption mechanism was declared in detail.

Materials and Reagents
The steel slag samples, used as a growth substrate for CNTs, were obtained from Baoshan Iron & Steel Co., Ltd. (Shanghai, China). X-ray fluorescence (XRF) analysis was utilized for detecting the chemical composition of the steel slag ( Table 1). The steel slag contained 33.98% Fe 2 O 3 , which is generally considered to be an appropriate catalyst for the growth of CNTs [27].
The lead nitrate (Pb(NO 3 ) 2 ), cupric nitrate (Cu(NO 3 ) 2 ), sodium hydroxide (NaOH) and hydrochloric acid (HCl) were of A.R. grade and were purchased from Shanghai Macklin Biochemical Technology Co., Ltd., Shanghai, China. Deionized water was used for both the preparation of the stock solutions and the dilution of the adsorption solutions into different concentrations. The initial pH values of the solutions were adjusted with HCl (0.1 M) or NaOH (0.1 M). Figure 1 presents a diagram of the CNTs' synthesis device, which consisted of a gas mass-flow controller and a horizontal tube furnace. The diameter of the quartz tube inside the furnace was 80 mm, and the length of the furnace heating zone was 200 mm. For synthesis of CNTs by the CVD method, acetylene, hydrogen and nitrogen were utilized as a carbon source, reducing gas and carrier gas, respectively. In this work, 10 g of wellground steel slag was uniformly loaded into a porcelain boat and subsequently placed in the heating zone of the furnace. The whole experimental process was as follows. First, nitrogen at a flow rate of 200 sccm was injected into the furnace to remove oxygen until the quartz tube was heated to the desired synthesis temperature. Then, nitrogen, hydrogen and acetylene were introduced into the furnace at a certain rate. After a certain synthesis time, the hydrogen and acetylene flows were turned off, while nitrogen was kept at a flow rate of 800 sccm until the quartz tube had cooled down to ambient temperature. Both the rate and time varied. The yield of carbon obtained could be calculated as follows:

Synthesis of the CNTs
where X is the yield of carbon obtained (wt%); M and M are the total mass of the synthetic product (g) and the mass of raw steel slag (g), respectively.
Nanomaterials 2022, 12, x FOR PEER REVIEW Figure 1. The schematic diagram of the carbon nanotubes' synthesis device.

Adsorption Study
In this work, adsorption experiments were performed in a series of 200 mL p ylene bottles with 20 mg of adsorbents and 100 mL of heavy metal ion solutions at v concentrations. The polyethylene bottles were placed in a water bath oscillator and at 150 rpm (revolutions per minute) for various adsorption times. The adsorption ca for heavy metal ions was evaluated by the following formula: where is the adsorption capacity (mg/g); and C represent the initial ion c tration (mg·L −1 ) and the ion concentration (mg·L −1 ) after equilibrium adsorption fo tain time (min), respectively; is the volume of the solution (L); is the amoun SS@CNTs (g).

Characterization
The yield and oxidation temperature of carbon products obtained by CVD w termined by thermogravimetric analysis (TGA, Netzsch STA 449 F3 Jupiter, Selb many) with a heating rate of 10 °C·min −1 and a temperature range of 30 to 1000 °C air atmosphere. The microstructural characteristics of synthetic CNTs were analyz ing a field emission scanning electron microscope (SEM, FEI NOVA NANOSEM 43 tham, Massachusetts, USA) at a working voltage of 10 kV. The diameters and thick of the CNTs were detected by a transmission electron microscope (TEM, JEM-210 kyo, Japan). The microscopic Raman spectrometer (HJY Lab RAM Aramis, Paris, F was employed to determine the crystallinity of CNTs with a spectra range of 400 t cm −1 . The particle sizes of products were investigated by a laser particle analyzer (L Horiba, Kyoto, Japan). The Brunauer-Emmett-Teller (BET) method was emplo characterize the specific surface area (SBET) and pore characters of the products. Th centrations of heavy metal irons were tested by an inductively coupled plasma emission spectrometer (ICP-OES).

Synthesis of CNTs on Steel Slag Particles by CVD
The growth of CNTs is affected by various factors. In this study, the synthes perature, growth time and acetylene flow rate were varied to study the effects of sy parameters on the growth of CNTs utilizing steel slag as the substrate.

Adsorption Study
In this work, adsorption experiments were performed in a series of 200 mL polyethylene bottles with 20 mg of adsorbents and 100 mL of heavy metal ion solutions at various concentrations. The polyethylene bottles were placed in a water bath oscillator and stirred at 150 rpm (revolutions per minute) for various adsorption times. The adsorption capacity for heavy metal ions was evaluated by the following formula: where q i is the adsorption capacity (mg/g); C 0 and C i represent the initial ion concentration (mg·L −1 ) and the ion concentration (mg·L −1 ) after equilibrium adsorption for a certain time (min), respectively; V is the volume of the solution (L); m is the amount of the SS@CNTs (g).

Characterization
The yield and oxidation temperature of carbon products obtained by CVD was determined by thermogravimetric analysis (TGA, Netzsch STA 449 F3 Jupiter, Selb, Germany) with a heating rate of 10 • C·min −1 and a temperature range of 30 to 1000 • C in an air atmosphere. The microstructural characteristics of synthetic CNTs were analyzed using a field emission scanning electron microscope (SEM, FEI NOVA NANOSEM 430, Waltham, Massachusetts, USA) at a working voltage of 10 kV. The diameters and thicknesses of the CNTs were detected by a transmission electron microscope (TEM, JEM-2100F, Tokyo, Japan). The microscopic Raman spectrometer (HJY Lab RAM Aramis, Paris, France) was employed to determine the crystallinity of CNTs with a spectra range of 400 to 4000 cm −1 . The particle sizes of products were investigated by a laser particle analyzer (LA960S, Horiba, Kyoto, Japan). The Brunauer-Emmett-Teller (BET) method was employed to characterize the specific surface area (S BET ) and pore characters of the products. The concentrations of heavy metal irons were tested by an inductively coupled plasma optical emission spectrometer (ICP-OES).

Synthesis of CNTs on Steel Slag Particles by CVD
The growth of CNTs is affected by various factors. In this study, the synthesis temperature, growth time and acetylene flow rate were varied to study the effects of synthesis parameters on the growth of CNTs utilizing steel slag as the substrate. SS@CNTs-45 and SS@CNTs-60, respectively). The flow rates of H 2 , C 2 H 2 and N 2 were set as 400, 200 and 800 sccm, respectively, with a synthesis temperature of 600 • C.
The results of the TG analysis are shown in Figure 2a and Table 2. The first region (460 to 670 • C) and the second region (670 to 710 • C) corresponded to the oxidization of carbon products and the decomposition of calcium carbonate (about 4 wt% loss), respectively [28,29]. The yield of carbon products on the steel slag particles increased as the growth time increased, reaching a maximum weight loss of 39% at 60 min. However, the increased rate of the yield was higher before 45 min and lower between 45 and 60 min, which was attributed to the rapid increase in precipitated carbon atoms before 45 min and the continuous decrease in active sites on the catalyst [30]. Of note, there was little difference in the initial oxidation temperature (T 0 ) of the carbon products under different times, while the final oxidation temperature (T f ) of the carbon products increased as the growth time increased. The degree of crystallization of the carbon products was basically related to the oxidation temperature. Carbonaceous species with higher crystallinity were more stable and, thus, had a higher oxidation temperature. A longer growth time ensured enough time for carbon atoms to deposit and diffuse on the substrate and then form fewer defective carbon products that had a higher oxidation temperature [31,32].
x FOR PEER REVIEW 5 of 22 30, SS@CNTs-45 and SS@CNTs-60, respectively). The flow rates of H2, C2H2 and N2 were set as 400, 200 and 800 sccm, respectively, with a synthesis temperature of 600 °C. The results of the TG analysis are shown in Figure 2a and Table 2. The first region (460 to 670 °C) and the second region (670 to 710 °C) corresponded to the oxidization of carbon products and the decomposition of calcium carbonate (about 4 wt% loss), respectively [28,29]. The yield of carbon products on the steel slag particles increased as the growth time increased, reaching a maximum weight loss of 39% at 60 min. However, the increased rate of the yield was higher before 45 min and lower between 45 and 60 min, which was attributed to the rapid increase in precipitated carbon atoms before 45 min and the continuous decrease in active sites on the catalyst [30]. Of note, there was little difference in the initial oxidation temperature (T0) of the carbon products under different times, while the final oxidation temperature (Tf) of the carbon products increased as the growth time increased. The degree of crystallization of the carbon products was basically related to the oxidation temperature. Carbonaceous species with higher crystallinity were more stable and, thus, had a higher oxidation temperature. A longer growth time ensured enough time for carbon atoms to deposit and diffuse on the substrate and then form fewer defective carbon products that had a higher oxidation temperature [31,32].   15  461  640  13  30  465  657  23  45  463  672  36  60 463 673 39 Figure 2b shows the Raman spectra of SS@CNTs samples. The D-peak in the range of 1250 to 1350 cm −1 corresponded to the disordered and defective character of the amorphous carbon, while the G-peak in the range of 1500-1600 cm −1 depended essentially on the stretching vibration of sp 2 hybridized carbon atoms observed in most of the graphite carbonaceous species. The ratio of D-peak to G-peak intensity (ID/IG) is often used as a probe of the graphitization degree of CNTs [33]. As shown in Figure 2b, the value of ID/IG decreased from 1.38 to 1.19 as the growth time increased from 15 to 60 min, implying that the graphitization degree of the CNTs improved as the time increased. In addition, a peak   Figure 2b shows the Raman spectra of SS@CNTs samples. The D-peak in the range of 1250 to 1350 cm −1 corresponded to the disordered and defective character of the amorphous carbon, while the G-peak in the range of 1500-1600 cm −1 depended essentially on the stretching vibration of sp 2 hybridized carbon atoms observed in most of the graphite carbonaceous species. The ratio of D-peak to G-peak intensity (I D /I G ) is often used as a probe of the graphitization degree of CNTs [33]. As shown in Figure 2b, the value of I D /I G decreased from 1.38 to 1.19 as the growth time increased from 15 to 60 min, implying that the graphitization degree of the CNTs improved as the time increased. In addition, a peak between 2600 and 3000 cm −1 was also related to the sp 2 hybridized carbon atoms, and its peak intensity also had a similar trend.

Time (min) T0 (°C) Tf (°C) Yield of Carbon (wt%)
The Scanning electron microscope (SEM) images of the SS@CNTs synthesized under different times are shown in Figure 3. The CNTs were coiled around the surface of the steel slag particles, especially in the steep sections. As the growth time increased, the quantity of the CNTs and amorphous carbon increased and decreased, respectively, while the homogeneousness of the CNTs improved as the synthesis time increased. Notably, the highest length-diameter ratio of CNTs appeared at the synthesis time of 45 min. Furthermore, the CNTs that synthesized for a longer time seemed to be larger, which was clearly observed in TEM images ( Figure 4). The diameters of the SS@CNTs samples were 10.3, 12.8, 17.5 and 21.3 nm, respectively, while the thicknesses were 1.6, 3.4, 4.2 and 6.2 nm, respectively. A longer growth time meant enough time for the deposition and diffusion of carbon atoms onto the particles, which increased the quantity, diameter, thickness and homogeneousness of the CNTs. Similar to the TG analysis, amorphous carbon existed in all of the samples, which was attributed to the heterogeneous and complex surfaces of the steel slag particles. The Scanning electron microscope (SEM) images of the SS@CNTs synthesized under different times are shown in Figure 3. The CNTs were coiled around the surface of the steel slag particles, especially in the steep sections. As the growth time increased, the quantity of the CNTs and amorphous carbon increased and decreased, respectively, while the homogeneousness of the CNTs improved as the synthesis time increased. Notably, the highest length-diameter ratio of CNTs appeared at the synthesis time of 45 min. Furthermore, the CNTs that synthesized for a longer time seemed to be larger, which was clearly observed in TEM images ( Figure 4). The diameters of the SS@CNTs samples were 10.3, 12.8, 17.5 and 21.3 nm, respectively, while the thicknesses were 1.6, 3.4, 4.2 and 6.2 nm, respectively. A longer growth time meant enough time for the deposition and diffusion of carbon atoms onto the particles, which increased the quantity, diameter, thickness and homogeneousness of the CNTs. Similar to the TG analysis, amorphous carbon existed in all of the samples, which was attributed to the heterogeneous and complex surfaces of the steel slag particles.     Figure 5 shows the particle size distribution and cumulative size distribution of SS@CNTs under different growth times. As shown in Figure 5a, the sizes of SS@CNT particles were in the range of 5-220 μm. When the growth time increased, the particle size distribution curves moved in parallel to the direction of larger particles. Similarly, from the cumulative distribution curves (Figure 5b), we found that the particle diameter corresponding to a 90% cumulative volume (D90) of SS@CNTs synthesized at 15, 30, 45 and 60 min were 71, 91, 101, 112 and 145 μm, respectively, indicating that a longer growth time produced larger particles. The BET surface area (SBET) and porosity characters are shown in Table 3. The SBET increased before 45 min and then decreased, while the total pore volume (Vt) and average pore diameter (DA) increased as the time increased. The highest value of the BET surface area was obtained at 45 min with 49.85 m 2 ·g −1 , indicating that more CNTs and other carbon products with a large surface area existed at a growth time of 45 min. The increasing trend of the Vt and DA was related to the diameter of CNTs and the quantity of the amorphous carbon [34,35]. Based on all of the results above, an appropriate growth time was 45 min.   Figure 5a, the sizes of SS@CNT particles were in the range of 5-220 µm. When the growth time increased, the particle size distribution curves moved in parallel to the direction of larger particles. Similarly, from the cumulative distribution curves (Figure 5b), we found that the particle diameter corresponding to a 90% cumulative volume (D90) of SS@CNTs synthesized at 15, 30, 45 and 60 min were 71, 91, 101, 112 and 145 µm, respectively, indicating that a longer growth time produced larger particles. The BET surface area (S BET ) and porosity characters are shown in Table 3. The S BET increased before 45 min and then decreased, while the total pore volume (V t ) and average pore diameter (D A ) increased as the time increased. The highest value of the BET surface area was obtained at 45 min with 49.85 m 2 ·g −1 , indicating that more CNTs and other carbon products with a large surface area existed at a growth time of 45 min. The increasing trend of the V t and D A was related to the diameter of CNTs and the quantity of the amorphous carbon [34,35]. Based on all of the results above, an appropriate growth time was 45 min.    Figure 6a and Table 4 present the TG analysis of SS@CNTs synthesized at various temperatures (500, 600, 700 and 800 °C corresponding with samples SS@CNT-500, SS@CNT-600, SS@CNT-700 and SS@CNT-800, respectively). Furthermore, the flow rates of H2, C2H2 and N2 were 400, 200 and 800 sccm, respectively, with a synthesis time of 45 min. According to the result, the yield of CNTs on steel slag particles increased first and then decreased. It is well known that the yield of CNTs is related to the growth rate, which can be explained by the Arrhenius equation [36]. A higher temperature promotes the decomposition of the carbon precursor, the diffusion and precipitation of carbon atoms and the catalyst activity, resulting in the improvement in the growth rate. The maximum yield of 36% was at 600 °C, indicating that this is a suitable temperature for CNTs to form onto steel slag particles. Moreover, according to Table 4, as the synthesis temperature increased, the initial and final oxidation temperatures significantly increased. This phenomenon can be explained by the decrease in the amorphous carbon and the increase in the crystallinity of carbon products at 600 °C and above.   Table 4 present the TG analysis of SS@CNTs synthesized at various temperatures (500, 600, 700 and 800 • C corresponding with samples SS@CNT-500, SS@CNT-600, SS@CNT-700 and SS@CNT-800, respectively). Furthermore, the flow rates of H 2 , C 2 H 2 and N 2 were 400, 200 and 800 sccm, respectively, with a synthesis time of 45 min. According to the result, the yield of CNTs on steel slag particles increased first and then decreased. It is well known that the yield of CNTs is related to the growth rate, which can be explained by the Arrhenius equation [36]. A higher temperature promotes the decomposition of the carbon precursor, the diffusion and precipitation of carbon atoms and the catalyst activity, resulting in the improvement in the growth rate. The maximum yield of 36% was at 600 • C, indicating that this is a suitable temperature for CNTs to form onto steel slag particles. Moreover, according to Table 4, as the synthesis temperature increased, the initial and final oxidation temperatures significantly increased. This phenomenon can be explained by the decrease in the amorphous carbon and the increase in the crystallinity of carbon products at 600 • C and above. The Raman spectra of SS@CNT samples are shown in Figure 6b. The values of ID/IG decreased from 1.37 to 1.13 as the synthesis temperature increased from 500 to 800 °C. When the synthesis temperature was below 600 °C, more amorphous carbon formed owing to the low catalyst activity and inefficient decomposition of acetylene. However, as the temperature increased to 800 °C, the value of ID/IG decreased to 1.13, implying that the  The Raman spectra of SS@CNT samples are shown in Figure 6b. The values of I D /I G decreased from 1.37 to 1.13 as the synthesis temperature increased from 500 to 800 • C. When the synthesis temperature was below 600 • C, more amorphous carbon formed owing to the low catalyst activity and inefficient decomposition of acetylene. However, as the temperature increased to 800 • C, the value of I D /I G decreased to 1.13, implying that the crystallinity of carbon products improved. Figure 7 shows the SEM images of SS@CNT samples. As the synthesis temperature increased, the length and quantity of CNTs increased first and then decreased, reaching the maximum at 600 • C, while the diameter of CNTs increased as the temperature increased. When the temperature was below 600 • C, fewer CNTs formed on the particles. Notably, as the temperature increased to 600 • C, the length and quantity of CNTs increased rapidly, and the highest aspect ratio of CNTs was observed. However, as the temperature continuously increased from 600 to 800 • C, the length of CNTs decreased, but the diameter increased, leading to the formation of thick and short CNTs. From the TEM result (Figure 8), we found that the diameter and wall thickness increased continually. The diameters of the samples were approximately 8.4, 17.5, 22.6 and 31.4 nm. The thicknesses of samples were approximately 2.2, 4.2, 6.5 and 6.7 nm, respectively. For a lower temperature, the activity of the catalyst as well as the diffusion and deposition rate of carbon were lower. As a result, carbon atoms were preferentially deposited on smaller particles with a higher catalytic activity to form thinner carbon nanotubes. As the temperature increased, the activity of catalyst particles increased, allowing the decomposition of carbon atoms on a larger surface of the catalyst particles, combined with the increased diffusion rate of carbon atoms on the substrate, which led to the formation of larger CNTs. An overly high temperature caused the catalyst to be inactive as a result of being covered by carbon layers, either graphitic or amorphous. Hence, shorter and thicker CNTs formed at a temperature above 600 • C [37].
The particle size distribution of the SS@CNTs are shown in Figure 9. With the increase in temperature, the particle sizes increased first and then decreased, and the values of D90 under different temperatures were 62, 112, 91 and 76 µm, respectively. This phenomenon can be attributed to the variation trend in the quantity and morphology of CNTs. Table 5 presents the BET surface area and porosity characteristics of the SS@CNTs samples. As the temperature increased, the maximum BET surface area and total pore volume were obtained at a growth temperature of 600 • C, indicating that the quantities of CNTs and other carbon products with a large surface area formed. Based on all of the results above, an appropriate synthesis temperature was at 600 • C. Nanomaterials 2022, 12, x FOR PEER REVIEW 10 of 22   can be attributed to the variation trend in the quantity and morphology of CNTs. Table 5 presents the BET surface area and porosity characteristics of the SS@CNTs samples. As the temperature increased, the maximum BET surface area and total pore volume were obtained at a growth temperature of 600 °C, indicating that the quantities of CNTs and other carbon products with a large surface area formed. Based on all of the results above, an appropriate synthesis temperature was at 600 °C.
(a) (b) Figure 9. The particle size distribution (a) and cumulative size distribution (b) of SS@CNTs synthesized at various temperatures.

Effect of the Acetylene Flow Rate
To investigate the effect of acetylene flow rate on the synthesis of SS@CNTs, several samples were prepared (sample SS@CNT-100, SS@CNT-200 and SS@CNT-300 corresponded with SS@CNTs synthetized under V C 2 H 2 :V N 2 = 100:800, V C 2 H 2 :V N 2 = 200:800 and V C 2 H 2 :V N 2 = 300:800, respectively) at 600 °C for 45 min. Figure 10a and Table 6 show the TGA result of SS@CNTs under different C2H2 flow rates. According to the results, as the C2H2 flow rate increased, the yield of carbon products increased. When the flow rate of C2H2 increased from 100 to 200 sccm, the yield of carbon products significantly increased, from 17 to 36 wt%. A high ratio of the carbon precursor flow rate to carrier gas flow rate provided sufficient contact time between the carbon precursor and catalyst particle and, hence, formed more carbon products. From Table 6, we can see that the samples were similar among the initial oxidation temperatures, while the final temperature increased as the flow rate of C2H2 increased. This indicated that carbon products with low crystallinity existed in all samples and a high acetylene flow rate seemed to improve the crystallinity.
The Raman spectra of the SS@CNTs samples is shown in Figure 10b. As the C2H2 flow rate increased, the value of ID/IG decreased first and then increased. The values of ID/IG in SS@CNTs-100 and SS@CNTs-300 were higher than those in SS@CNTs-200, suggesting that

Effect of the Acetylene Flow Rate
To investigate the effect of acetylene flow rate on the synthesis of SS@CNTs, several samples were prepared (sample SS@CNT-100, SS@CNT-200 and SS@CNT-300 corresponded with SS@CNTs synthetized under V C 2 H 2 :V N 2 = 100:800, V C 2 H 2 :V N 2 = 200:800 and V C 2 H 2 :V N 2 = 300:800, respectively) at 600 • C for 45 min. Figure 10a and Table 6 show the TGA result of SS@CNTs under different C 2 H 2 flow rates. According to the results, as the C 2 H 2 flow rate increased, the yield of carbon products increased. When the flow rate of C 2 H 2 increased from 100 to 200 sccm, the yield of carbon products significantly increased, from 17 to 36 wt%. A high ratio of the carbon precursor flow rate to carrier gas flow rate provided sufficient contact time between the carbon precursor and catalyst particle and, hence, formed more carbon products. From Table 6, we can see that the samples were similar among the initial oxidation temperatures, while the final temperature increased as the flow rate of C 2 H 2 increased. This indicated that carbon products with low crystallinity existed in all samples and a high acetylene flow rate seemed to improve the crystallinity. As shown in the SEM results (Figure 11), narrow and short CNTs presented in the SS@CNTs-100, while more amorphous carbon and short and thick CNTs existed in SS@CNTs-300. In contrast, quantities of long and uniform CNTs with a high aspect ratio presented in SS@CNTs-200. From the TEM images (Figure 12), we found that the diameter and wall thickness of CNTs increased as the C2H2 flow rate increased. The diameters of CNTs in SS@CNTS-100, SS@CNTs-200 and SS@CNTs-300 were approximately 7, 13, and 41 nm, respectively. The wall thicknesses were approximately 1.2, 2.2 and 14.5 nm, respectively. As other researchers pointed out, the low flow rate of C2H2 caused a short contact time between the carbon precursor and catalyst particles; thus, the decomposition of C2H2 and the diffusion of carbon atoms were insufficient for the further growth of CNTs. However, the overly high C2H2 flow rate meant an excess carbon supply, which led to the increase in amorphous carbonaceous products and other impurities on the catalyst particles as well as on the surface of grown CNTs. Furthermore, the catalyst particles were poisoned by the amorphous carbon and, thus, formed short and thick CNTs [38,39].  The Raman spectra of the SS@CNTs samples is shown in Figure 10b. As the C 2 H 2 flow rate increased, the value of I D /I G decreased first and then increased. The values of I D /I G in SS@CNTs-100 and SS@CNTs-300 were higher than those in SS@CNTs-200, suggesting that the quantity of amorphous carbonaceous species and the defects of CNTs were lower under a C 2 H 2 flow rate of 200 sccm. The crystallinity of carbon products decreased either above or below this flow rate.
As shown in the SEM results ( Figure 11), narrow and short CNTs presented in the SS@CNTs-100, while more amorphous carbon and short and thick CNTs existed in SS@CNTs-300. In contrast, quantities of long and uniform CNTs with a high aspect ratio presented in SS@CNTs-200. From the TEM images (Figure 12), we found that the diameter and wall thickness of CNTs increased as the C 2 H 2 flow rate increased. The diameters of CNTs in SS@CNTS-100, SS@CNTs-200 and SS@CNTs-300 were approximately 7, 13, and 41 nm, respectively. The wall thicknesses were approximately 1.2, 2.2 and 14.5 nm, respectively. As other researchers pointed out, the low flow rate of C 2 H 2 caused a short contact time between the carbon precursor and catalyst particles; thus, the decomposition of C 2 H 2 and the diffusion of carbon atoms were insufficient for the further growth of CNTs. However, the overly high C 2 H 2 flow rate meant an excess carbon supply, which led to the increase in amorphous carbonaceous products and other impurities on the catalyst particles as well as on the surface of grown CNTs. Furthermore, the catalyst particles were poisoned by the amorphous carbon and, thus, formed short and thick CNTs [38,39].  Figure 13 shows the particle size distribution and the cumulative size distribution of the samples. The particle size increased as the C2H2 flow rate increased. The values of D90 in samples SS@CNTs-100, SS@CNTs-200 and SS@CNTs-300 were approximately 91, 113 and 129 μm, respectively. It seemed that the particle size was related to the quantity of carbon products on the steel slag particles. More carbon products led to larger SS@CNT particles. According to the BET surface areas and porosity characters of CNTs (Table 7), as the C2H2 flow rate increased, the BET surface area increased first and then decreased, reaching the maximum value in the SS@CNT-200 sample. In addition, as the C2H2 flow rate increased, the total pore volume increased, while the average pore diameter decreased first and then increased. As observed in the SEM images, a larger diameter in CNTs meant a larger average pore diameter in the samples. The increase in the total pore volume was the result of the increase in the entire carbon product quantity. Based on all of the results above, we concluded that 200 sccm was an appropriate C2H2 flow rate in this study.  Figure 13 shows the particle size distribution and the cumulative size distribution of the samples. The particle size increased as the C2H2 flow rate increased. The values of D90 in samples SS@CNTs-100, SS@CNTs-200 and SS@CNTs-300 were approximately 91, 113 and 129 μm, respectively. It seemed that the particle size was related to the quantity of carbon products on the steel slag particles. More carbon products led to larger SS@CNT particles. According to the BET surface areas and porosity characters of CNTs (Table 7), as the C2H2 flow rate increased, the BET surface area increased first and then decreased, reaching the maximum value in the SS@CNT-200 sample. In addition, as the C2H2 flow rate increased, the total pore volume increased, while the average pore diameter decreased first and then increased. As observed in the SEM images, a larger diameter in CNTs meant a larger average pore diameter in the samples. The increase in the total pore volume was the result of the increase in the entire carbon product quantity. Based on all of the results above, we concluded that 200 sccm was an appropriate C2H2 flow rate in this study.  Figure 13 shows the particle size distribution and the cumulative size distribution of the samples. The particle size increased as the C 2 H 2 flow rate increased. The values of D90 in samples SS@CNTs-100, SS@CNTs-200 and SS@CNTs-300 were approximately 91, 113 and 129 µm, respectively. It seemed that the particle size was related to the quantity of carbon products on the steel slag particles. More carbon products led to larger SS@CNT particles. According to the BET surface areas and porosity characters of CNTs (Table 7), as the C 2 H 2 flow rate increased, the BET surface area increased first and then decreased, reaching the maximum value in the SS@CNT-200 sample. In addition, as the C 2 H 2 flow rate increased, the total pore volume increased, while the average pore diameter decreased first and then increased. As observed in the SEM images, a larger diameter in CNTs meant a larger average pore diameter in the samples. The increase in the total pore volume was the result of the increase in the entire carbon product quantity. Based on all of the results above, we concluded that 200 sccm was an appropriate C 2 H 2 flow rate in this study.

Adsorption Kinetics
A series of experiments were performed at 25 °C with an initial solution pH of 6.5. The contact time was in a range from 15 to 300 min (Figure 14a). The adsorption rate was significantly high during the first 15 min, and then gradually slowed down, reaching equilibrium at approximately 90 min. The first rapid adsorption stage suggested that there were abundant adsorption sites on the adsorbent surface, which reacted with heavy metal ions quickly. The maximum equilibrium adsorption capacities of SS@CNTs to Pb(II) and Cu(II) were 427.26 mg·g −1 and 132.79 mg·g −1 , respectively. For further study of the adsorption process, the pseudo-first-order kinetic and the pseudo-second-order kinetic were carried out, which can be described by the formulas: where is the equilibrium adsorption capacity for heavy metal ions (mg·g −1 ); is the adsorption capacity (mg·g −1 ) at time (t, min); and are the pseudo-first-order kinetic and the pseudo-second-order kinetic reaction rate constants, respectively.
The linear fitting diagrams and the corresponding parameters of the kinetic models are shown in Figure 14b,c and Table 8, respectively. The best fitting model was determined by comparing the correlation coefficients (R 2 ) and the differences between the values and the experimental ones. Based on the R 2 values, the pseudo-second-order kinetic model described the adsorption process more accurately than the pseudo-first-order kinetic; moreover, the values of , calculated from the pseudo-second-order kinetic

Adsorption Kinetics
A series of experiments were performed at 25 • C with an initial solution pH of 6.5. The contact time was in a range from 15 to 300 min (Figure 14a). The adsorption rate was significantly high during the first 15 min, and then gradually slowed down, reaching equilibrium at approximately 90 min. The first rapid adsorption stage suggested that there were abundant adsorption sites on the adsorbent surface, which reacted with heavy metal ions quickly. The maximum equilibrium adsorption capacities of SS@CNTs to Pb(II) and Cu(II) were 427.26 mg·g −1 and 132.79 mg·g −1 , respectively. For further study of the adsorption process, the pseudo-first-order kinetic and the pseudo-second-order kinetic were carried out, which can be described by the formulas: where q e is the equilibrium adsorption capacity for heavy metal ions (mg·g −1 ); q t is the adsorption capacity (mg·g −1 ) at time (t, min); k a and k b are the pseudo-first-order kinetic and the pseudo-second-order kinetic reaction rate constants, respectively.

Adsorption Isotherms
Adsorption isotherms are of great help for adsorption studies. In this study, two models (i.e., Langmuir and Freundlich) were used to fit the data obtained from adsorption equilibrium experiments, which were performed at temperatures of 15, 25 and 45 °C for 3 h. The initial concentrations of Pb(II) and Cu(II) were in a range from 80 to 160 mg·L −1 and 40 to 160 mg·L −1 , respectively.
The Langmuir adsorption isotherm equation is: where is the maximum adsorption capacity of heavy metal ions by the adsorbent (mg·g −1 ); is the Langmuir equilibrium constant, which is concerned with the bonding stability between adsorbents and adsorbates (L·mg −1 ); is the equilibrium concentration [42].
The Freundlich adsorption isotherm equation is: where is the Freundlich constant, and is a constant determined by the adsorption intensity of the adsorbent [43].
The isothermal adsorption curves of Pb(II) and Cu(II) at the temperature of 15, 25 and 45 °C were shown in Figure 15a,b. The linear fitting charts are shown in Figure 15c-f. The corresponding parameters (i.e., , , , and ) and the values of R 2 were obtained by regressive computation of the linearized Langmuir and Freundlich isotherm models, which are presented in Table 9. From them, it is clear that Pb(II) and Cu(II) were well adsorbed by SS@CNTs. By comparing values of R 2 , the Langmuir model (R 2 > 0.99) explained the removal of Pb(II) and Cu(II) by SS@CNTs better than the Freundlich model, implying the uniformly distributed adsorption sites as well as the monolayer adsorption of heavy metal ions on the surface of SS@CNTs. The maximum adsorption amounts calculated from the linearized Langmuir isotherm increased with the increasing temperature, which was consistent with the experimental results. In addition, all of the values of were in the range of 0 to 1, indicating the favorable adsorption of Pb(II) and Cu(II) The linear fitting diagrams and the corresponding parameters of the kinetic models are shown in Figure 14b,c and Table 8, respectively. The best fitting model was determined by comparing the correlation coefficients (R 2 ) and the differences between the q e values and the experimental ones. Based on the R 2 values, the pseudo-second-order kinetic model described the adsorption process more accurately than the pseudo-first-order kinetic; moreover, the values of q e , calculated from the pseudo-second-order kinetic model, were also approximately the same as those obtained by experiments. The agreement of the pseudosecond-order process with the experimental results implied that the removal process was a chemical adsorption [40,41].

Adsorption Isotherms
Adsorption isotherms are of great help for adsorption studies. In this study, two models (i.e., Langmuir and Freundlich) were used to fit the data obtained from adsorption equilibrium experiments, which were performed at temperatures of 15, 25 and 45 • C for 3 h. The initial concentrations of Pb(II) and Cu(II) were in a range from 80 to 160 mg·L −1 and 40 to 160 mg·L −1 , respectively.
The Langmuir adsorption isotherm equation is: where q m is the maximum adsorption capacity of heavy metal ions by the adsorbent (mg·g −1 ); K L is the Langmuir equilibrium constant, which is concerned with the bonding stability between adsorbents and adsorbates (L·mg −1 ); C e is the equilibrium concentration [42]. The Freundlich adsorption isotherm equation is: where K F is the Freundlich constant, and n is a constant determined by the adsorption intensity of the adsorbent [43].
The isothermal adsorption curves of Pb(II) and Cu(II) at the temperature of 15, 25 and 45 • C were shown in Figure 15a,b. The linear fitting charts are shown in Figure 15c-f. The corresponding parameters (i.e., q m , K L , R L , K F and n) and the values of R 2 were obtained by regressive computation of the linearized Langmuir and Freundlich isotherm models, which are presented in Table 9. From them, it is clear that Pb(II) and Cu(II) were well adsorbed by SS@CNTs. By comparing values of R 2 , the Langmuir model (R 2 > 0.99) explained the removal of Pb(II) and Cu(II) by SS@CNTs better than the Freundlich model, implying the uniformly distributed adsorption sites as well as the monolayer adsorption of heavy metal ions on the surface of SS@CNTs. The maximum adsorption amounts calculated from the linearized Langmuir isotherm increased with the increasing temperature, which was consistent with the experimental results. In addition, all of the values of R L were in the range of 0 to 1, indicating the favorable adsorption of Pb(II) and Cu(II) onto SS@CNTs. Based on the parameters of the Freundlich model, the values of k F of both Pb(II) and Cu(II) increased as the temperature increased, suggesting that the uptake of these two ions was heightened at a high temperature. Furthermore, the values of n varied from 6.588 to 8.178 for Pb(II) and from 3.047 to 3.949 for Cu(II), which indicates that the removal of both ions occurred due to the fact of adsorption on the heterogeneous surface [44,45]. SS@CNTs. Based on the parameters of the Freundlich model, the values of of both Pb(II) and Cu(II) increased as the temperature increased, suggesting that the uptake of these two ions was heightened at a high temperature. Furthermore, the values of varied from 6.588 to 8.178 for Pb(II) and from 3.047 to 3.949 for Cu(II), which indicates that the removal of both ions occurred due to the fact of adsorption on the heterogeneous surface [44,45].

Adsorption Thermodynamics
To investigate the thermodynamic characteristics of the adsorption process, a batch of experiments were performed at the temperatures of 15, 25 and 45 • C for 3 h with an initial pH of 6.5. The thermodynamic parameters, such as free energy (∆G 0 ), enthalpy (∆H 0 ) and entropy (∆S 0 ), were defined as follows: where K L (L·mol −1 ) and R (8.314 J·mol −1 ·K −1 ) are the Langmuir equilibrium constant and the gas constant, respectively. The ∆H 0 and ∆S 0 were obtained from the slope and intercept of the linear plot (ln K L versus 1 T ), which are shown in Figure 16a.

Adsorption Thermodynamics
To investigate the thermodynamic characteristics of the adsorption process, a batch of experiments were performed at the temperatures of 15, 25 and 45 °C for 3 h with an initial pH of 6.5. The thermodynamic parameters, such as free energy (∆ 0 ), enthalpy (∆ 0 ) and entropy (∆ 0 ), were defined as follows: where (L·mol −1 ) and (8.314 J·mol −1 ·K −1 ) are the Langmuir equilibrium constant and the gas constant, respectively. The ∆ 0 and ∆ 0 were obtained from the slope and intercept of the linear plot (ln versus 1 ), which are shown in Figure 16a. Table 10 shows the thermodynamics parameters of the adsorption process at different temperatures. The values of ∆ 0 were negative and decreased as the temperature increased, suggesting that the adsorption of heavy metals onto SS@CNTs was spontaneous, and a high temperature was more favorable for the adsorption reaction. Moreover, the positive values of ∆ 0 also imply a spontaneous adsorption as well as the random character of the solution-adsorbate interface. From the positive values of ∆ 0 , it can be concluded that the adsorption processes were endothermic; thus, the adsorption reaction was more likely to occur at a high temperature.    Table 10 shows the thermodynamics parameters of the adsorption process at different temperatures. The values of ∆G 0 were negative and decreased as the temperature increased, suggesting that the adsorption of heavy metals onto SS@CNTs was spontaneous, and a high temperature was more favorable for the adsorption reaction. Moreover, the positive values of ∆S 0 also imply a spontaneous adsorption as well as the random character of the solution-adsorbate interface. From the positive values of ∆H 0 , it can be concluded that the adsorption processes were endothermic; thus, the adsorption reaction was more likely to occur at a high temperature. The pH value affects not only the chemical state of the adsorbate ions but also ionization of functional groups on the adsorbent surface. To explore the influence of pH on the elimination of Pb(II) and Cu(II), a batch of experiments were performed under a pH range of 2 to 7 at 25 • C for 3 h. The results are shown in Figure 16b. Obviously, the uptake of these two metal ions by SS@CNTs increased as the pH values increased. Increasing pH leads to ionization of functional groups on the graphite surface that consequently encourages the interaction of metal cations with these appeared anions grafted to the surface. Moreover, the decrease in hydrogen ions competing with heavy metal ions also makes a contribution to the increase in heavy metal removal. Note that the adsorption rate increased rapidly after the pH value reached 4.5, which is related to the point of zero charge (pH PZC ) of the absorbent surface [46]. Based on the change of the zeta potential of the SS@CNTs with the pH varying from 1 to 8, the pH PZC of the absorbent surface was estimated at approximately 4.2. When the pH value was higher than pH PZC , the adsorbent surface was negatively charged so as to increase the likelihood of attracting positively charged heavy metal ions [2].

Comparison and Adsorption Mechanism
By comparison of the uptake of lead and copper ions by different materials reported in the literature, which is shown in Table 11, the elimination of Pb(II) and Cu(II) by SS@CNTs was highly efficient. The adsorption, ion exchange, surface coordination and precipitation presented the main adsorption types of the elimination of Pb(II) and Cu(II) by SS@CNTs. During the adsorption process, in terms of steel slag, the calcium silicate and the free calcium oxide (f-CaO) continuously reacted with water to form calcium silicate hydrate (C-S-H) and calcium hydroxide (Ca(OH) 2 ), respectively, and then released calcium ions (Ca 2+ ) and hydroxide ions (OH -) [47]. The C-S-H was of great help in the removal of metal ions because of its high specific surface area as well as the ion exchange reaction between heavy metal ions and Ca 2+ . The OHrelease not only increased the pH of the solution but also formed hydroxide precipitation with heavy metal ions, which further improved the adsorption. Moreover, the hydroxyl group on the surface of the particles formed negatively charged adsorption sites that attracted the metal ions onto the surface of SS@CNTs by electrostatic interaction. The corresponding reaction formulas are as follows [24,25,48,49]: 2CaO·SiO 2 + nH 2 O → xCaO·SiO 2 ·yH 2 O + (2 − x)Ca(OH) 2 (10) The CNTs had large specific surface areas and did well in the adsorption of heavy metals. The main type of adsorption was physical absorption, such as electrostatic interaction, which has been discussed by many other researchers [50,51]. The combination of CNTs and steel slag increased the overall specific surface area, the quantity of adsorption sites and the types of adsorption, which greatly enhanced the adsorption capacity for heavy metal ions. Considering the low cost and significant adsorption capacity, the SS@CNTs' composite adsorbent has good application prospect in the elimination of heavy metals, such as lead and copper, from wastewater.

Conclusions
In this study, an efficient and low-cost CNT-steel slag composite adsorbent was directly synthesized by means of the CVD method and applied to remove lead and copper from an aqueous solution. In this way, the CNTs on steel slag particles were evenly dispersed in the aqueous solution with the dispersion of the particles so as to obtain better contact with the metal ions. The optimum growth time, synthesis temperature and C 2 H 2 flow rate for the SS@CNTs synthesis were 45 min, 600 • C and 200 sccm, respectively, and the CNTs showed a larger length-diameter ratio, a higher specific surface area and better morphology compared to the CNTs synthesized under other conditions. The SS@CNTs adsorbent did well in the removal of Pb(II) and Cu(II) with maximum equilibrium adsorption capacities of 427.26 mg·g −1 and 132.79 mg·g −1 , respectively, at 25 • C for an adsorption time of 90 min. The pseudo-second-order model and isotherms of the Langmuir model provided the most accurate fitting with the adsorption process of Pb(II) and Cu(II). The adsorption thermodynamics results indicated that the adsorption process was spontaneous and endothermic. SS@CNTs had a large specific surface area and abundant adsorption sites and types, owing to the combination of the advantages of CNTs and steel slag. In comparison to other adsorbent materials, SS@CNTs had a better adsorption performance for Pb(II) and Cu(II). In summary, SS@CNTs that utilize a steel industrial byproduct for raw material have good application prospects in the elimination of heavy metals, such as lead and copper, from wastewater.