1 Introduction

Industrial, agricultural, and domestic activities inevitably produce large amounts of wastewater [1,2,3]. Pollutants in the water can mainly be divided into three categories: microorganisms, organics, and inorganics. Heavy metal ions, as the major part of inorganic pollutants, have aroused lots of concern due to their toxicities to ecological and biological systems [4,5,6,7,8]. The United States Environmental Protection Agency (EPA) has limited the levels of various metal ions in drinking water. Table 1 lists some of them and the corresponding effects on human health [9].

Table 1 EPA maximum contaminant levels in drinking water and corresponding health effects
Full size table

Water pollution due to metal ions is still a serious problem and great challenge throughout the world. A tremendous amount of research focuses on this area, and many methods have been developed, such as adsorption [10, 11], ion exchange [12, 13], membrane filtration [14, 15], electro-chemical precipitation [16, 17], reverse osmosis [18, 19], and flocculation [20, 21]. Among these methods, adsorption is most widely used to remove metal ions from wastewater due to its simple, economical, and highly efficient characteristics [22, 23]. Lots of adsorbents including carbon materials [24, 25], metal oxides [26, 27], clays [28, 29], zeolites [30, 31], biomass [32, 33], and polymers [34, 35] have been explored.

Carbon nanomaterials, some of which are shown in Fig. 1, are the most commonly employed adsorbents for metal ion removal from wastewater owing to their large specific surface area and concentrated pore distribution [36, 37]. Moreover, they exhibit a great mechanical property and excellent structural stability under extreme conditions (e.g., high temperature and strong acidic/alkaline conditions) [38,39,40]. However, it is often difficult to disperse them in an aqueous environment and recover them after adsorption due to their hydrophobicity and small size [41,42,43]. In addition, some carbon nanomaterials, such as graphene, tend to aggregate decreasing their surface area and surface energy, which strongly decreases their adsorption performance [44,45,46]. To overcome these disadvantages, much attention has been paid to decorating the carbon nanomaterials. Various modified carbon nanoadsorbents have been developed because of the widespread functionalization of carbon nanomaterials.

Fig. 1
figure 1

Some carbon nanostructures. Reprinted with permission from Ivyspring International Publisher [36]

Full size image

In this review, the uses of modified carbon nanoadsorbents for metal ion removal from water are focused on. Different types of carbon-based nanoadsorbents and their adsorption performances, common methods to modify carbon nanomaterials, various adsorption isotherm and kinetic models, diverse adsorption mechanisms, and practical applications of the modified carbon nanoadsorbents to environmental water samples are covered. Furthermore, some issues about current research are discussed. For example, the adsorbents with such characteristics as high selectivity, extreme sensitivity, or none toxicity are rarely reported. It is hoped that this review can help to promote the development of carbon-based nanoadsorbents and the treatment for water pollution.

2 Overview

Based on their different architectures, carbon nanomaterials include zero-dimensional carbon nanoparticles and carbon nanospheres, one-dimensional carbon nanotubes (CNTs) and carbon nanofibers, two-dimensional graphene and carbon nanofabrics, and three-dimensional carbon nanoclusters and carbon nanofoams. A variety of adsorbents involving them are thus developed.

2.1 Graphene-based nanoadsorbents

Graphene possesses a two-dimensional structure consisting of sp2 hybridized carbon atoms covalently bonded in a honeycomb or hexagonal lattice with only one atomic thickness [47,48,49,50]. Due to its superior thermal, mechanical, electronic, and chemical properties, graphene and its derivatives can be applied to many fields, such as electronics, energy, sensors, and composites [51,52,53,54]. Moreover, since it has a huge surface area, large delocalized π-electron system, and tunable chemical properties, graphene and its derivatives can be used to effectively adsorb metals from wastewater [55,56,57]. Graphene as the nanoadsorbent mainly includes three forms: pristine graphene, graphene oxide (GO), and reduced graphene oxide (rGO) [58]. GO is an oxidized form of graphene, containing various oxygen-containing groups, such as epoxide, carbonyl, carboxyl, and hydroxyl groups [59,60,61,62,63]. Since the oxygen groups break the double bonds holding the carbon atoms together, GO loses its electrical conductivity. However, these oxygen groups cause GO to become polar and easy to disperse in water [64,65,66]. rGO is the reduction product of GO, which can be prepared through thermal, chemical, or electrical treatments of GO. Compared to pristine graphene, rGO contains more such defects as residual oxygen and other heteroatoms and has a lower conductivity.

Xu et al. [67] successfully prepared polyacrylamide grafted graphene (PAM-g-graphene) through γ-ray treatment of graphite oxide and acrylamide, which was characterized by UV-Vis absorption spectroscopy, XRD, XPS, FTIR, TGA, and AFM. The as-prepared PAM-g-graphene had a thickness of 2.59 nm, N2-BET-specific surface area of 128 m2/g, and grafted PAM chains of 24.2 wt%. In contrast, the N2-BET specific surface area of pristine graphite oxide and the thickness of its single-layer sheets were 46.4 m2/g and 1.30 nm, respectively. A superior performance for Pb(II) adsorption was exhibited by the PAM-g-graphene. Figure 2 a, b shows that the adsorption equilibrium could be reached in 30 min, and the adsorption kinetics was well described by the pseudo-second order model. Langmuir and Freundlich isotherm models were employed to illustrate the adsorption isotherms, as shown in Fig. 2 c, d. The maximum adsorption capacity determined by the Langmuir isotherm model was 819.67 mg/g, which was respectively 20 times and 8 times higher than that of graphene nanosheets and CNTs. The one-step synthesis of PAM-g-graphene exhibited a bright prospect for heavy metal removal from water.

Fig. 2
figure 2

a Effect of the contact time on the adsorption of Pb(II) onto PAM-g-graphene (adsorbent 0.1 g/L; initial Pb(II) concentration 45 mg/L; pH 6; T 293 K); b adsorption kinetics of Pb(II) onto PAM-g-graphene fitted by pseudo-second-order model; adsorption isotherm of Pb(II) onto PAM-g-graphene fitted by c Langmuir model and d Freundlich model (adsorbent 0.1 g/L; initial Pb(II) concentration 5–100 mg/L; pH 6; T 293 K). Reprinted with permission from Elsevier [67]

Full size image

As shown in Fig. 3, L-cystine functionalized exfoliated GO (EGO) was synthesized and used for Hg(II) adsorption [68]. First, graphite was oxidized and exfoliated to prepare GO. Then GO was treated with thionyl chloride (SOCl2), and acylchloride groups were formed due to the reaction between carboxyl groups of the GO and SOCl2. After that, L-cystine was added, and its amine groups reacted with acylchloride groups to produce acylamide groups. In this way, L-cystine was successfully grafted onto EGO. Nitrogen, oxygen, and sulfur-containing groups on the L-cystine functionalized EGO could effectively bind Hg(II), and an outstanding adsorption performance was observed. Compared with unmodified graphite, L-cystine functionalized EGO had a much higher adsorption capacity under the same conditions, which were 79.36 and 12.4 mg/g, respectively. Moreover, there was no significant interference from other metal ions when they were at low concentrations. Thiourea could be employed to accelerate the process of desorption, and after four adsorption-desorption cycles, the adsorption efficiency began to decrease.

Fig. 3
figure 3

Preparation process of L-cystine functionalized EGO and its interactions with Hg(II). Reprinted with permission from Royal Society of Chemistry [68]

Full size image

Radionuclides can cause serious environmental problems. Wen et al. [69] prepared three-dimensional hierarchical flower-like GO-hydroxyapatite (GO-HAp) nanocomposites for Sr(II) removal. The GO was synthesized through a modified Hummers method, and the HAp grew on the GO nanosheets via a biomimetic method. The process of HAp nucleation and growth was shown in Fig. 4. To evaluate the adsorption ability of GO-HAp, effects of contact time, solution pH, coexisting cations, GO-HAp content, and Sr(II) initial concentrations were investigated. The adsorption equilibrium could be quickly reached within 2 h, and the removal kinetics followed the pseudo-second-order model. The GO-HAp could keep a high removal percentage in a wide pH range, more than 76% even at pH 2–4. Cd(II) and Pb(II) had a slight influence on the adsorption of Sr(II), while Mg(II), Al(III), and Na(I) had almost no effects. As more GO-HAp were added to the solution, the adsorption percentage of Sr(II) increased swiftly, whereas its distribution coefficient gradually decreased. Both the Langmuir and Freundlich isotherms could well describe the adsorption behaviors of Sr(II). A maximum adsorption capacity of 702.18 mg/g was obtained from the Langmuir model, which was higher than most other adsorbents, almost two times that of bare HAp and nine times that of GO. Two possible mechanisms for Sr(II) adsorption were proposed, namely ion exchange between Ca(II) and Sr(II) and complex compound formation of Sr(II) with HAp active surface sites. The GO-HAp showed great potential applications for radiostrontium pollution clean-up.

Fig. 4
figure 4

The process of HAp nucleation and growth on GO. SEM images of the GO after a 12 h, b 1 day, c 3 days, and d 7 days of immersion in a 1.5 × simulated body fluid aqueous solution. Reprinted with permission from Royal Society of Chemistry [69]

Full size image

Yang et al. [70] employed a new adsorbent, lignosulfonate-GO-polyaniline (LS-GO-PANI) ternary nanocomposite, to adsorb Pb(II) from water. GO was first fabricated from expansible graphite via an airtight oxidation method, and then LS-GO-PANI was prepared through an in situ polymerization of aniline in the presence of LS and GO in a HCl aqueous solution, as shown in Fig. 5. Compared with PANI and GO-PANI, LS-GO-PANI ternary nanocomposite exhibited a higher adsorption capacity due to the existence of a synergistic effect among the functional groups on it. It was believed that the amino groups in PANI units enhanced the coordinate ability of sulfonic groups on LS chains and carboxyl groups on GO nanosheets for Pb(II). An optimal adsorption pH of 5.0 was observed for the LS-GO-PANI, and 98.3% of its maximum adsorption capacity could be reached in 4 h. Pb(II) concentrations and LS-GO-PANI amounts had an effect on the adsorption capacity and adsorptivity, which could further utilized to enhance the adsorption of Pb(II).

Fig. 5
figure 5

Preparation process of LS-GO-PANI ternary nanocomposites. Reprinted with permission from American Chemical Society [70]

Full size image

2.2 Carbon nanotube-based nanoadsorbents

Carbon nanotubes (CNTs) are allotropes of carbon that are long, thin, and cylindrical, about 1–3 nm in diameter and hundreds to thousands of nanometers long. They are a kind of graphite material and can be viewed as rolled up graphene sheets [71, 72]. CNTs are mainly divided into single-walled CNTs (SWCNTs) and multiwalled CNTs (MWCNTs) [73, 74]. SWCNTs have only a single layer of graphene, while MWCNTs are made up of multiple rolled layers of graphene. CNTs can be treated by acid, such as nitric acid and sulfuric acid, and potassium permanganate, to form oxidized CNTs [75,76,77,78]. Compared to CNTs, oxidized CNTs become polar and easy to disperse in water. CNTs and graphene are quite similar in many aspects, like their compositions, electrical, thermal, and optical properties [79]. Therefore, CNTs can be applied to such areas as electronics, energy, and sensors [80,81,82,83]. They are also an effective adsorbent for metal ion removal from water [84, 85].

Wang et al. [86] prepared amidoxime-grafted MWCNTs (AO-g-MWCNTs) to adsorb U(VI) from nuclear industrial effluents. Oxidized MWCNTs were first treated by N2 plasma and then grafted with acrylonitrile (AN) to produce AN-g-MWCNTs. The cyano groups on the AN-g-MWCNTs were converted to amidoxime groups by a reaction with the hydroxylamine hydrochloride. Compared to oxidized MWCNTs, AO-g-MWCNTs had a rougher surface and more compact stacking morphology. N2 adsorption-desorption experiments showed that the BET-specific surface area of oxidized MWCNTs decreased from 91.31 to 72.59 m2/g after plasma grafting. The adsorption behaviors of U(VI) onto AO-g-MWCNTs were systematically investigated. As the initial pH of the U(VI) solution increased from 1 to 4.5, the distribution coefficient of U(VI) gradually became larger. It only needed about 1 h to reach the adsorption equilibrium, and the adsorption kinetics could be best described by the pseudo-second-order model (R 2 = 0.9997). The Langmuir isotherm model revealed that AO-g-MWCNTs had a much higher maximum adsorption capacity than oxidized MWCNTs, which meant that AO could significantly enhance the adsorption ability of MWCNTs. Temperature also affected the adsorption of U(VI), and higher temperatures were favorable. The thermodynamic studies indicated that the adsorption process was spontaneous and U(VI) was more likely to be adsorbed by AO-g-MWCNTs than by oxidized MWCNTs. AO-g-MWCNTs exhibited much higher selectivity towards U(VI) over other coexisting metal ions, such as Ni(II), Zn(II), Cs(I), and Ba(II), and showed great removal performance for U(VI) from seawater. The adsorption mechanism of AO-g-MWCNTs was believed to be that U(VI) combined oxygen and nitrogen atoms of amidoxime groups to form complexes to achieve the removal of U(VI), as shown in Fig. 6. The AO functionalized MWCNTs, as a quite effective adsorbent, showed great potential applications in practice.

Fig. 6
figure 6

Probable mechanism for adsorption of U(VI) by AO-g-MWCNTs. Reprinted with permission from Elsevier [86]

Full size image

Tian et al. [87] synthesized poly(m-phenylenediamine)-coated iron oxide/acid-oxidized MWCNTs (PmPD/Fe3O4/o-MWCNTs) magnetic nanocomposites via in situ oxidative polymerization for Cr(VI) removal for the first time. The morphology of the nanocomposites at different stages in the synthesis process was shown in Fig. 7. BET calculation revealed that Fe3O4/o-MWCNTs and PmPD/Fe3O4/o-MWCNTs had a much lower pore volume, specific surface area, and pore size than o-MWCNTs, and the reason was believed to be that Fe3O4 nanoparticles and PmPD blocked the pore entrances on the o-MWCNTs. The adsorption capacities of PmPD/Fe3O4/o-MWCNTs, raw-MWCNTs, o-MWCNTs, and Fe3O4/o-MWCNTs were in the order of PmPD/Fe3O4/o-MWCNTs > Fe3O4/o-MWCNTs > o-MWCNTs > raw-MWCNTs. Effects of the initial pH of the solution on the adsorption of Cr(VI) onto PmPD/Fe3O4/o-MWCNTs were investigated, and the results showed that the adsorption capacity decreased as the pH increased from 2 to 11. Langmuir, Freundlich, and Temkin isotherm models were respectively used to describe the adsorption behaviors of Cr(VI), and the Langmuir model fitted the adsorption data best. As the temperature increased from 273 to 333 K, the maximum adsorption capacity calculated from the Langmuir model increased from 219.8 to 346.0 mg/g. Adsorption dynamic studies showed that both the pseudo-first-order model and pseudo-second-order model could be employed to simulate the adsorption kinetics, and the pseudo-second-order equation gave a relatively higher correlation coefficient (R 2 > 0.99). Thermodynamic studies indicated that the adsorption of Cr(VI) onto PmPD/Fe3O4/o-MWCNTs was a spontaneous and endothermic process. After adsorption, PmPD/Fe3O4/o-MWCNTs could be easily separated through an external magnetic field, and its adsorption capacity decreased by less than 48% after five consecutive recycles. Both PmPD groups and Fe3O4 particles played an important role in the adsorption of Cr(VI), and the adsorption mechanism included physical and chemical adsorptions. In conclusion, PmPD/Fe3O4/o-MWCNT was a good candidate for wastewater treatment.

Fig. 7
figure 7

TEM images of a, b pure Fe3O4 nanoparticles, c, d Fe3O4/o-MWCNTs nanocomposites, and e, f PmPD/Fe3O4/o-MWCNTs nanocomposites at different magnifications. Reprinted with permission from Royal Society of Chemistry [87]

Full size image

Magnetic hydroxypropyl chitosan/oxidized MWCNT (MHC/o-MWCNT) composites were successfully fabricated and employed as the adsorbent for Pb(II) removal from an aqueous solution, as shown in Fig. 8 [88]. Magnetic Fe3O4 nanoparticles were first prepared from FeCl3·6H2O, and then MWCNTs were oxidized by a mixture of concentrated sulfuric acid and nitric acid to produce o-MWCNTs. Afterwards, magnetic Fe3O4 nanoparticles and o-MWCNTs were sequentially added to the hydroxypropyl chitosan solution, and MHC/o-MWCNTs composites were assembled. The MHC/o-MWCNTs composites showed an impressive adsorption performance with an adsorption capacity of 116.3 mg/g. Moreover, they exhibited a rapid response to the external magnetic field, and the separation time was as fast as 3 min after adsorption. Sips isotherm model was found more suitable than Langmuir, Freundlich, and Dubinin-Radushkevich isotherm models to describe the adsorption behaviors of Pb(II) onto MHC/o-MWCNTs composites. The thermodynamic parameters, such as free energy − 2.304 KJ/mol, enthalpy 39.03 KJ/mol, and entropy 138.7 J/(mol·K), illustrated that the adsorption process was endothermic and spontaneous. The MHC/o-MWCNTs composites were an ideal adsorbent to remove Pb(II) from water.

Fig. 8
figure 8

Preparation of MHC/o-MWCNT composites and their application for Pb(II) removal. Reprinted with permission from Elsevier [88]

Full size image

2.3 Other carbon-based nanoadsorbents

Graphene and CNTs are the two most widely studied carbon nanoadsorbents. Besides them, there are many other carbon nanomaterials that are used as nanoadsorbents, such as carbon nanofabrics, nanofibers, nanocables, and nanofloral clusters. These carbon nanoadsorbents often exhibit some distinctive characteristics due to their different architectures, which may bring about outstanding adsorption performance, economical costs, and easy recovery [89,90,91,92].

Zhu et al. [93] employed conventional and microwave-assisted annealing methods, respectively, to synthesize mesoporous magnetic carbon fabric nanocomposites for Cr(VI) removal. For comparison, Fig. 9a shows the removal percentage of different adsorbents in neutral solution. Both MH and MN (magnetic carbon fabric nanocomposites prepared by microwave annealing in a 5% H2-Ar and N2 atmosphere, respectively) exhibited complete Cr(VI) removal at an initial Cr(VI) concentration of 1.5 mg/L, while CN (magnetic carbon fabric nanocomposites fabricated via conventional annealing under a N2 atmosphere) only had 68.7% of Cr(VI) removed, and even lower values were reported for cotton fabrics (CottF, 21.3%) and carbon fabrics (CarbF, 30.7%). The MN had a higher removal percentage than MH when the initial concentration of Cr(VI) was 4.0 mg/L. Figure 9b illustrated the effect of initial Cr(VI) concentration on the removal efficiency of MN. A maximum removal percentage of 100% was achieved for the initial Cr(VI) concentration of 1.5, 2.0, and 2.5 mg/L. However, the removal percentage decreased with Cr(VI) concentration further increasing. Figure 9c indicates the Cr(VI) removal performance at different MN concentrations. The removal percentage increased with the increase of MN loadings. The maximum removal percentage of 100% could be achieved with an MN loading of 2.0 g/L. The pH effect on the removal efficiency of Cr(VI) using MN was revealed in Fig. 9d. The MN showed stronger removal ability when the pH was lower than 3. With an increasing pH value, the removal percentage decreased (about 78.4% at pH 11). The MN had an adsorption capacity of 3.74 mg/g, which was much higher than carbon-coated magnetic nanoparticles (1.52 mg/g) and graphene nanocomposites (1.03 mg/g).

Fig. 9
figure 9

a The Cr(VI) removal performance of different materials (adsorbent 1.0 g/L; treating time 10 min; pH 7); b effect of the initial Cr(VI) concentration on the Cr(VI) removal performance (MN 1.0 g/L; treating time 10 min; pH 7); c Cr(VI) removal percentage at different MN concentrations (initial Cr(VI) concentration 4.0 mg/L; treating time 10 min; pH 7); d effect of the solution pH on the Cr(VI) removal efficiency (MN 1.0 g/L; initial Cr(VI) concentration 4.0 mg/L; treating time 10 min). Reprinted with permission from Royal Society of Chemistry [93]

Full size image

Conventional micro−/nanostructured adsorbents often suffer from unfavorable aggregation, which will lead to overlap and coverage of numerous adsorption sites. As a result, their adsorption efficiency seriously decreased. A snowflake-shaped magnetic ZnO@SiO2@Fe3O4/C micro−/nanostructure was prepared, as shown in Fig. 10 [94]. First, a snowflake-shaped porous ZnO micro−/nanostructure was fabricated through a facile hydrothermal route followed by a sintering process. Then, the ZnO micro−/nanostructures were coated with SiO2 layers via the Stöber method. Next, magnetic Fe3O4/C composite coatings were grown onto the ZnO@SiO2 micro−/nanostructures through a solvothermal approach. The prepared ZnO@SiO2@Fe3O4/C micro−/nanostructures could keep a relatively large space between themselves, which significantly reduced the aggregation of micro−/nanostructures and the overlap of adsorption sites. Compared with magnetic ZnO@SiO2@Fe3O4/C microspheres prepared by the similar procedures, the snowflake-shaped magnetic ZnO@SiO2@Fe3O4/C micro−/nanostructures exhibited a much higher adsorption efficiency. The maximum adsorption capacities for Pb(II) and As(V) were 94.3 and 23.6 mg/g, respectively, while for the magnetic ZnO@SiO2@Fe3O4/C microspheres, those were only 54.6 and 17.2 mg/g. The mechanism for the enhancement was demonstrated from the special biomimetic structure.

Fig. 10
figure 10

Preparation process of snowflake-shaped magnetic ZnO@SiO2@Fe3O4/C micro−/nanostructures. Reprinted with permission from the Royal Society of Chemistry [94]

Full size image

Magnetic carbon-iron nanoadsorbents fabricated by carbonizing cellulose and reducing Fe3O4 nanoparticles or Fe(NO3)3 (denoted as MC-O and MC-N, respectively) had demonstrated great Cr(VI) removal ability [95]. The SEM images (Fig. 11 a, b) showed that they presented spherical particles with an average diameter of ~ 100 and 250 nm for MC-O and MC-N, respectively. A higher portion of zero-valence iron (ZVI) and a larger specific surface area were observed for the MC-N than the MC-O. Both MC-O and MC-N exhibited a great Cr(VI) removal performance. For example, 4.0 mg/L Cr(VI) neutral solution could be purified by 2.5 g/L MC-O and MC-N within 10 min, and 1000 mg/L Cr(VI) could be completely removed from the solution in 10 min at pH 1.0. The equilibrium adsorption data could be well fitted for Langmuir isotherm model (Fig. 11c). Meanwhile, the Cr(VI) removal by both MC-O and MC-N was highly pH dependant (Fig. 11d), and the removal percentage decreased with the pH value increasing in solutions. MC-N and MC-O revealed removal capacities of 327.5 and 293.8 mg/g at pH 1.0, respectively, which were much higher than many other adsorbents, such as magnetic chitosan (55.8 mg/g), magnetic carbon fabricated by rice husk (30.96 mg/g), and polypyrrole-polyaniline nanofibers (227.22 mg/g). For the removal mechanisms, it was believed that Cr(VI) was first reduced to Cr(III) by the ZVI, and then the Cr(III) was removed from the solution through precipitation and adsorption.

Fig. 11
figure 11

SEM images of a MC-O and b MC-N; c Cr(VI) adsorption isotherm fitted by Langmuir model (adsorbent dosage 50.0 mg; volume 20.0 mL; initial Cr(VI) concentration 1.0–10.0 mg/L; treating time 30 min; pH 7); d effect of the solution pH on the Cr(VI) removal performance (adsorbent dosage 50.0 mg; volume 20.0 mL; initial Cr(VI) concentration 4.0 mg/L; treating time 10 min). Reprinted with permission from the Royal Society of Chemistry [95]

Full size image

Han et al. [96] successfully fabricated [C/Fe3O4]@C electricity-magnetism-adsorption trifunctional coaxial nanocables by carbonization of the electrospun [polyacrylonitrile/ferric acetylacetonate]@polyacrylonitrile coaxial nanocables. The as-prepared [C/Fe3O4]@C coaxial nanocables had obvious core-shell structure with a core diameter of 125 nm and shell thickness of 82 nm. Magnetic Fe3O4 nanoparticles were scattered in core of carbon matrix and displayed a positive effect on electrical conductivity. Efficient adsorption of Cu(II) from an aqueous solution was observed using [C/Fe3O4]@C coaxial nanocables as the adsorbent. Cui et al. [97] synthesized hybrid γ-Fe2O3/carbon hollow nanospheres with a predominant orientation (1 1 1) plane of γ-Fe2O3 and rich oxygen-containing functional groups on carbon via a one-step hydrothermal method. The hybrid γ-Fe2O3/carbon hollow nanospheres exhibited high adsorption capacities and rapid removal abilities for Pb(II) and Cr(VI). They could be easily recovered through an external magnetic field after adsorption. Mamun et al. [98] prepared carbon nanofibers (CNFs) by chemical vapor deposition method in which C2H2 was employed as carbon source and Ni(II) catalyst impregnated oil palm kernel shell-based powdered-activated carbon (PAC) was used as fixed substrate for the growth of CNFs. The PAC-CNFs porous nanocomposites were found to be good for adsorption of Pb(II). The adsorption kinetics and isotherm were carefully investigated, and all the results showed that the PAC-CNFs porous nanocomposites had potential for Pb(II) removal from an aqueous solution.

2.4 Comparison

Table 2 summarizes the adsorption data of some modified carbon nanomaterials that were recently reported. Their maximum adsorption capacities are compared, and the corresponding adsorbed metal ions are illustrated.

Table 2 Adsorption capacity comparison of different carbon nanoadsorbents
Full size table

It can be seen from Table 2 that graphene and CNTs are the two most often used carbon nanoadsorbents. In fact, most modified carbon nanoadsorbents involve graphene and CNTs composites. Besides that, polymers, especially those with electron-donating or alkaline groups, are increasingly applied to various carbon nanoadsorbents. They have many advantages over other materials, such as tailored molecular weight, chain topology, and functional groups [152,153,154]. Moreover, they are usually nontoxic as a result of their large molecular size, which will dramatically reduce the secondary pollution [155, 156]. The plentiful active groups of polymers can significantly enhance the adsorption performance, and more possible ways are provided to achieve the desorption and recovery of adsorbents in an easy manner with the advent of stimuli-responsive polymers [157,158,159]. In a word, it becomes more and more popular utilizing polymers to modify carbon nanoadsorbents.

3 Analysis and discussion

Based on the above summary, some common modification methods, various adsorption isotherm and kinetic models, and diverse adsorption mechanisms can be concluded.

3.1 Modification methods

The removal process of metal ions from wastewater mainly includes three steps: adsorption, desorption, and recovery. Carbon nanomaterials are always modified to optimize their performances in these areas by changing their morphologies or molecular structures. Some commonly used modification methods can be found by analyzing papers broken into the categories of magnetic carbon nanomaterials, electron-donating group grafted-carbon nanomaterials, mixed or filled carbon nanomaterials, and carbon nanomaterials with modified morphologies.

3.1.1 Magnetic carbon nanomaterials

Adsorbents should ideally be well-dispersed in water during the adsorption process and be easy to aggregate during the recovery process. However, it seems there are contradictions between these two processes because usually the better the dispersion of an adsorbent is, the more difficult the separation of it will be. Adsorbent particles with a smaller size are often easier to form a uniform solution than those with a larger size, but they are usually harder to be separated by such methods as filtration, centrifugation, and gravity sedimentation. Magnetic separation is an effective way to solve the above problem. It is a physical, easy to operate, economical, and environmentally friendly process. The adsorbents can be easily separated by applying an external magnetic field after introducing some magnetic materials into them. These magnetic materials often not only facilitate the recovery of adsorbents but also improve their adsorption ability. For graphene-based nanoadsorbents, they can even inhibit the aggregation of graphene nanosheets. Some frequently used magnetic materials are Fe3O4, γ-Fe2O3, and MFe2O4 (M = Zn, Mg, Co, Cu, Ni, Mn, etc.) [100, 101, 103, 104, 151].

Tan et al. [137] synthesized cobalt ferrite/MWCNTs (CoFe2O4/MWCNTs) magnetic hybrids by a hydrothermal method. The surface area and pore volume of CoFe2O4/MWCNTs were much larger than those of CoFe2O4. Compared to MWCNTs and CoFe2O4, CoFe2O4/MWCNTs showed a far higher adsorption capacity for U(VI). U(VI)-loaded CoFe2O4/MWCNTs could be easily separated from the solution by using a magnet. Chella et al. [100] fabricated graphene manganese ferrite (MnFe2O4-G) nanocomposite through a modified Hummer’s method and solvothermal process for the removal of Pb(II) and Cd(II). The existence of MnFe2O4 nanoparticles not only improved the adsorption capacity of graphene but also enhanced its antibacterial activity. Fe3O4 nanoparticles with an average size of 17 nm were successfully synthesized and decorated onto the surface of GO via ultrasound-assisted precipitations [114]. The as-prepared GO-Fe3O4 hybrid exhibited high adsorption efficiency for Co(II) and Sr(II), especially for Co(II) arisen from its close similarity to Fe(II) in the magnetite.

3.1.2 Electron-donating group grafted carbon nanomaterials

Since metal ions are positively charged, an effective method to enhance the adsorption ability of carbon nanomaterials is grafting electron-donating groups onto them, such as nitrogen, oxygen, sulfur, or chlorine-containing groups. These electron-donating groups can be viewed as bases and metal ions belong to acids based on Lewis acid-base theory. Their interactions include electrostatic interaction, ion exchange, complex formation, and redox reaction [160, 161]. Using this method not only improves the adsorption rate and capacity but also optimizes the dispersity of carbon nanomaterials in water due to the stronger hydrophilicity of grafted atoms in comparison with carbon atoms. The adsorption selectivity can also be enhanced by introducing some specific groups with high selectivity.

Based on the attraction of N atoms in amine groups to heavy metal ions which arises from their long pair electrons, Tan et al. [117] employed L-tryptophan to aminate GO to investigate its removal performance for Cu(II) and Pb(II). A greatly enhanced adsorption capacity was obtained. After three recycle loops, the L-tryptophan functionalized GO (GO/LTrp) exhibited less than a 5% decrease in adsorption capacity. Li et al. [136] for the first time synthesized dithiocarbamate groups functionalized MWCNT (DTC-MWCNT) to selectively adsorb Cd(II), Cu(II), and Zn(II) from an aqueous solution by reaction of oxidized MWCNTs with ethylenediamine and carbon disulfide. Since Cd(II), Cu(II), and Zn(II) were considered soft acids and dithiocarbamate was a soft base, the DTC-MWCNT exhibited high maximum adsorption capacities. To be more specific, they were 167.2, 98.1, and 11.2 mg/g for Cd(II), Cu(II), and Zn(II), respectively.

3.1.3 Mixed or filled carbon nanomaterials

Each adsorbent has its own characteristics, strengths, and weaknesses. If different adsorbents are added together in a proper way, a new adsorbent will be obtained possessing a performance superior to any individual adsorbent. In addition, dispersing active fillers into the matrix of adsorbents is another quite effective method to enhance their various properties, such as the mechanical strength and chemical stability. These fillers act like the additives in plastics, and some of them include metal oxides, polymers, and biomass. Sui et al. [124] employed polyethylenimine-modified magnetic GO (GO/Fe3O4/PEI) nanocomposites to remove Cu(II) from an aqueous solution. Fe3O4 nanoparticles first grew on GO sheets, and then the GO/Fe3O4 was mixed with PEI to obtain GO/Fe3O4/PEI nanocomposites. Superior Cu(II) removal performance was observed, which was attributed to the large surface area of GO, superparamagnetism of Fe3O4, and extraordinary complex ability of PEI. The GO and GO/Fe3O4/PEI had adsorption capacities of 89.32 and 157.48 mg/g, respectively, which clearly indicated the dominant role of PEI and Fe3O4. After 5 cycles of regeneration, the GO/Fe3O4/PEI nanocomposites could keep a removal efficiency as high as 84%. Based on the high ion-exchange capability of MgAl-layered double-hydroxide (LDH), strong adsorption ability of mesoporous carbon, and easy recovery of Fe3O4 nanoparticles, Zhang et al. [150] designed and developed novel Fe3O4@C@MgAl-LDH nanoparticles by chemical self-assembly methods. They could adsorb as much as 152.0 mg/g of Cr(VI) at 40 °C and pH 6.0. Besides the excellent adsorption capacity, they exhibited great reusability, which made them very suitable for the removal of metal ions from wastewater.

3.1.4 Carbon nanomaterials with modified morphologies

Besides to modify the compositions, another way to meet our requirements for some properties of carbon nanoadsorbents is designing and customizing their morphologies. For example, an increase in the surface area and porosity of carbon nanomaterials usually brings about the improvement in the adsorption rate and capacity.

Lei et al. [130] developed a novel three-dimensional (3D) GO foam/Fe3O4 nanocomposite (GOF/Fe3O4) and evaluated its adsorption performance for Cr(VI). Microwave-plasma chemical vapor deposition techniques were employed to synthesize the free-standing 3D graphene foam on the growth substrate of nickel foam. GOF/Fe3O4 was formed after the graphene foam was oxidized and functionalized with Fe3O4 nanoparticles through a simple coprecipitation method. The 3D structure provided GOF/Fe3O4 with an ultra-high specific surface area of 574.2 m2/g. Benefiting from that, a maximum adsorption capacity of 258.6 mg/g and equilibrium adsorption rate of 20 min were obtained, which markedly outperformed the performance of reported 2D graphene-based adsorbents and many other conventional adsorbents. Sankararamakrishnan et al. [142] used composite nanofloral clusters of CNTs and activated alumina for removal of Cd(II). CNTs were grown over Fe- and Ni-doped activated alumina by chemical vapor deposition and washed with acid to produce nanofloral clusters. The optimum adsorption condition for Cd(II) was in the pH range of 7–9, and the produced nanofloral clusters had the maximum adsorption capacity of 229.9 mg/g at pH 7.5, which was far higher than that of CNTs, oxidized CNTs, and amino functionalized CNTs. The nanofloral clusters as a promising candidate for wastewater treatment provided new ways to design effective adsorbents.

In practice, many modified carbon nanomaterials have changes in both morphologies and molecular structures. Per actual needs, they are designed and prepared to own some special or enhanced properties. Although there are many ways to enhance the adsorption abilities of carbon nanomaterials, such as changing their sizes, porosities, functional groups, and shapes, most of them aim to increase the contact between adsorbates and adsorption sites, which is one of the foundations for the development of carbon-based nanoadsorbents.

3.2 Adsorption isotherm, kinetics, and mechanisms

A variety of new modified carbon nanoadsorbents have been developed in the past few years. They may have different structures and properties. However, there are a few common models that can be employed to describe the adsorption of metal ions onto them.

3.2.1 Adsorption isotherm and kinetics

Adsorption isotherm and kinetics are often used to describe the adsorption process, mechanism, and performance. Specifically, adsorption isotherm gives information about adsorption capacities, types of adsorption systems, surface properties of adsorbents, and relationships between adsorbents and adsorbates [162, 163]; information about adsorption rates, rate-limiting steps, and diffusion processes of adsorbates can be given by adsorption kinetics. Tables 3 and 4 listed some usual adsorption isotherm and kinetic models.

Table 3 Adsorption isotherm models
Full size table
Table 4 Adsorption kinetic models
Full size table

Langmuir and Freundlich isotherm models are the two most often used isotherm models. The Langmuir isotherm assumes monolayer adsorption on a homogeneous surface, while the Freundlich isotherm assumes multilayer adsorption on a heterogeneous surface. The linear form of Langmuir equation is:

$$ \frac{C_{\mathrm{e}}}{Q_{\mathrm{e}}}=\frac{C_{\mathrm{e}}}{Q_{\mathrm{m}}}+\frac{1}{K_{\mathrm{L}}{Q}_{\mathrm{m}}} $$

where Q e is the equilibrium adsorption capacity of adsorbent, C e is the equilibrium concentration of adsorbate, Q m is the maximum adsorption capacity corresponding to complete monolayer coverage on the adsorbent surface, and K L is a constant related to the energy of adsorption. The feasibility of the adsorption process is determined by R L, which is given as [171,172,173]:

$$ {R}_{\mathrm{L}}=\frac{1}{1+{K}_{\mathrm{L}}{C}_0} $$

where C 0 is the initial concentration of adsorbate, and R L is known as the separation factor and is dimensionless. If R L is between 0 and 1, a favorable isotherm is represented, which means an effective interaction between the adsorbent and adsorbate. If R L is greater than 1, an unfavorable isotherm is implied. If R L equals 1, a linear isotherm is meant, and if R L equals 0, an irreversible isotherm is indicated.

The linear form of Freundlich isotherm can be expressed as follows:

$$ \log {\mathrm{Q}}_{\mathrm{e}}=\log {K}_{\mathrm{F}}+\frac{1}{n}\log {C}_{\mathrm{e}} $$

where Q e is the equilibrium adsorption capacity of adsorbent, C e is the equilibrium concentration of adsorbate, and K F and n are Freundlich constants related to the adsorption capacity and intensity, respectively. A favorable adsorption has an n between 1 and 10; a normal L-type isotherm is represented when 1/n < 1, which means an effective interaction between the adsorbent and adsorbate; if 1/n > 1, a co-adsorption is implied [174,175,176,177].

Zhou et al. [164] prepared sponge-like polysiloxane-GO gel via a simple one-step sol-gel method to remove Pb(II) and Cd(II) from wastewater. Dubinin-Radushkevish isotherm was employed to confirm that chemisorption occurred. Temkin isotherm assumes that the heat of adsorption decreases linearly rather than logarithmic with coverage. Instead of pure GO, Verma et al. [120] utilized ammonia-modified GO to enhance its adsorption selectivity towards U(VI), and the adsorption capacity was in good agreement with the Temkin model. Redlich-Peterson isotherm is a hybrid of the Langmuir and Freundlich isotherms. Najafabadi et al. [122] fabricated a chitosan/GO composite nanofibrous adsorbent through electrospinning for removal of Cu(II), Pb(II), and Cr(VI) from an aqueous solution. It was found that the Redlich-Peterson isotherm model fitted better than the Langmuir and Freundlich models. Sips isotherm is also called Langmuir-Freundlich isotherm. It is a combination of the Langmuir and Freundlich isotherms. At low adsorbate concentrations, it reduces to the Freundlich isotherm, while at high concentrations, it approaches the monolayer adsorption capacity like the Langmuir isotherm. A poly(sodium acrylate)-GO (PSA–GO) double network hydrogel adsorbent was prepared by Xu et al. and used to adsorb Cd(II) and Mn(II) [125]. The adsorption process could be best described by the Sips model. Henry isotherm is the simplest adsorption isotherm in that the equilibrium adsorption capacities are proportional to the adsorbate concentrations. It is typically taken as valid for low surface coverages. Guo et al. [105] synthesized amino functionalized magnetic graphene composites and studied their performance for adsorbing Cr(VI), Pb(II), Hg(II), Cd(II), and Ni(II) from an aqueous solution. The adsorption of Cr(VI) and Pb(II) was well fitted with the Henry model.

The most commonly used adsorption kinetic models are pseudo-first-order and pseudo-second-order models. Their linear equation forms are given as follows:

$$ {\displaystyle \begin{array}{l}\log \left({Q}_e-{Q}_t\right)=\log \left({Q}_e\right)-\left(\frac{k_1}{2.303}\right)t\\ {}\frac{t}{Q_{\mathrm{t}}}=\frac{1}{k_2{Q}_{\mathrm{e}}^2}+\left(\frac{1}{Q_{\mathrm{e}}}\right)t\end{array}} $$

where Q e and Q t are the adsorption capacities at equilibrium time and time t, respectively; k 1 and k 2 are the pseudo-first-order rate constant and pseudo-second-order rate constant, respectively. Pseudo-second-order model hypothesizes that the rate-limiting step is chemical adsorption involving valence forces through sharing or exchange of electrons between the adsorbent and adsorbate.

Tan et al. [167] prepared highly ordered layered GO membranes with larger interlayer spacing by an induced directional flow method and used them for removal of Cu(II), Cd(II), and Ni(II) from an aqueous solution. The kinetic data was consistent with intraparticle diffusion model, which meant that adsorption of those adsorbates onto the GO membranes included both external and intra-particle diffusion, and the intra-particle diffusion was not the only rate-limiting step. A magnetic biochar was synthesized by Wang et al. to adsorb As(V) from an aqueous solution, and Elovich model gave a good simulation result, which suggested that multiple interaction mechanisms or processes might control the adsorption of As(V) [169]. Cui et al. [123] fabricated magnetic carbonate hydroxyapatite/GO (M-CHAP/GO) and studied its Pb(II) removal abilities. Pseudo-first-order, pseudo-second-order, Elovich, intraparticle diffusion, and Bangham equations were applied to the description of adsorption kinetics, respectively. The pseudo-second-order model fitted best, and therefore, the main rate determining step was chemisorptions. In other words, the adsorption rate was controlled by chemical process through electronic exchange or chemical reactions between the M-CHAP/GO and Pb(II).

3.2.2 Interactions between adsorbents and adsorbates

It’s important to understand the mechanism for adsorption of metals by adsorbents, which plays a significant role in designing more effective adsorbents. The interactions between adsorbents and adsorbates mainly include electrostatic interaction, ion exchange, complex formation, and redox reaction [160, 161].

Yang et al. [134] utilized magnetic Prussian blue/GO (PB/Fe3O4/GO) nanocomposites to adsorb radioactive Cs(I) in water, which were synthesized by an in situ controllable method. It was found that the concentration of hydrogen ions obviously increased after adsorption, and the increased value was greater than the reduced amount of Cs(I). Thus, the mechanism for adsorption of Cs(I) onto PB/Fe3O4/GO may be H+-exchange and/or ion trapping. Zhou et al. [141] synthesized chitosan-functionalized MWCNT/CoFe2O4-NH2 (MNP-CTS) hybrid material for the removal of Pb(II). After the adsorption of Pb(II), the peak of N-H bond in the FTIR spectra of MNP-CTS shifted from 1558 to 1575 cm−1, which was attributed to the formation of stable complexes between Pb(II) and nitrogen in amine groups. Moreover, the peaks of C-O groups at 1071 and 1191 cm−1 shifted to 1043 and 1079 cm−1, respectively, and it was thought to be caused by the chelating and ion exchange between Pb(II) and O-H groups. Dinda et al. [99] polymerized 2,6-diamino pyridine on GO surfaces via mutual oxidation-reduction techniques to prepare sulfuric acid-doped poly diaminopyridine/graphene (G-PDAP) composite. The as-prepared P-DAP was investigated as an adsorbent to remove high concentration of toxic Cr(VI) from water. It was confirmed that the mechanisms for removal of Cr(VI) mainly included reduction of Cr(VI) to Cr(III) at low pH and anion exchange between mobile dopant and chromate ions at high pH levels. Cui et al. [135] successfully anchored ethylenediaminetetraacetic acid (EDTA) on Fe3O4 nanoparticles functionalized GO to obtain EDTA functionalized magnetic GO (GO-Fe3O4-EDTA) for the first time. The synthesized GO-Fe3O4-EDTA displayed excellent adsorption performance for Pb(II), Hg(II), and Cu(II). EDTA was believed to have a strong coordination interaction with metal ions. In addition, ζ potential analysis illustrated that the surface of GO-Fe3O4-EDTA was electronegative and the oxygenic functional groups on GO surface were deprotonated, which indicated that electrostatic attraction may be another reason for Pb(II), Hg(II), and Cu(II) binding to the GO-Fe3O4-EDTA.

4 Practical application

The environmental water samples are far more complex than that prepared in the laboratory due to some known or unknown substances, such as various metals, organics, and microorganisms. That an adsorbent performs very well in the laboratory cannot ensure its applicability to real water. Normally, the performance of an adsorbent decreases to some extent when it is used in practice. However, many modified carbon nanomaterials still show excellent adsorption ability in spite of those interfering substances in real water, which obviously demonstrates their practicality.

Chen et al. [133] attached polyethyleneimine (PEI) to the GO through an amidation reaction between the amine groups of PEI and the carboxyl groups of GO, which could not only effectively prevent the agglomeration of GO nanosheets but also greatly enhance its adsorption performance for metal ions. The prepared PEI-GO composite showed a much higher maximum uptake capacity for Cr(VI) than many other conventional adsorbents, such as ethylenediamine-functionalized Fe3O4, PVP-modified activated carbon, mesoporous TiO2, polypyrrole/Fe3O4 nanocomposite, hierarchical porous carbon, and chitosan-Fe(III) complex. Moreover, a removal percentage of 96.4% was observed after the PEI-GO at a concentration of 0.01 g/L was applied to the electroplating wastewater (Zhangzhou electroplating factory, China) spiked with 11.0 mg/L Cr(VI). It was evident that coexisting ions in electroplating wastewater had a weak influence on the adsorption of Cr(VI) onto PEI-GO. Aliyari et al. [55] reported the synthesis and application of diethylenetriamine-functionalized magnetic GO nanocomposites (GO-Fe3O4-DETA) as an adsorbent for the simultaneous separation and preconcentration of Pb(II) and Cd(II) from real water and vegetable samples prior to their determination by flame atomic absorption spectrometry. Sea water (the Caspian sea), river water (the Babolrud river), well water (a well in Tehran), treated lettuce, celery, and potato samples were investigated, respectively. More than 94% of Pb(II) and Cd(II) in those real samples were recovered and the relative standard deviations were less than 3.5%, which clearly demonstrated the practical applicability of GO-Fe3O4-DETA nanocomposites. Shaheen et al. [76] achieved the selective adsorption of Au(III) from an aqueous solution containing such interfering ions as Zn(II), Mn(II), Pb(II), and Cd(II) using oxidized MWCNTs (o-MWCNTs) that were prepared by oxidation of MWCNTs. The o-MWCNTs were also applied to real environmental water samples, including tap water collected from the laboratory, ground water and sea water collected from Jeddah city (Saudi Arabia), treated wastewater collected from the wastewater treatment station at King Abdulaziz University, and finally drinking water (Aquafina bottled water). All those real samples were spiked with Au(III) at the concentrations of 2, 5, and 10 mg/L, respectively, and the o-MWCNTs were kept at 0.4 mg/mL in them. The results exhibited that almost all the Au(III) were removed, which indicated that the o-MWCNTs were quite reliable, feasible, and suitable for the selective adsorption of Au(III) from real water samples. Ensafi et al. [73] oxidized and modified MWCNTs with concentrated HNO3 and thiolated cyanuric acid, respectively, and the treated MWCNTs were employed as adsorbents for the separation and preconcentration of Cd(II) and Pb(II) in various water samples, such as river water from the Zayandeh-Roud river (Isfahan, Iran), industrial wastewater (Mobarake Steel Complex, Isfahan, Iran), tap water, radiator manufacturing wastewater, and rice sample solution. The results showed a quite high recovery percentage of Cd(II) and Pb(II). Zhou et al. [75] used MWCNTs as the adsorbent and sodium diethyldithiocarbamate as the chelating agent for the simultaneous enrichment of Ni(II), Co(II), and Hg(II) in river water, sewage water, and factory water, respectively, and satisfied results were achieved.

5 Conclusions and prospects

It is of the utmost importance that water is clean and safe to use. However, water pollution is a serious problem around the world, especially in the developing countries. Millions of people are suffering from the threats of unhealthy water. Adsorption technologies are regarded as some of the most prevailing devices to mitigate water challenges, since they make it possible to supply clean water while being low cost and having a low energy consumption. Carbon-based nanomaterials as adsorbents have gained much attention in recent years due to their high stable structures, unique properties, and outstanding removal performances. Different types of modified carbon nanoadsorbents and their adsorption performances, common methods to modify carbon nanomaterials, various adsorption isotherm and kinetic models, diverse adsorption mechanisms, and practical applications of the modified carbon nanoadsorbents are covered. Currently, most research focuses on graphene- and CNT-based nanoadsorbents. Adsorption behaviors are best described by Langmuir or Freundlich isotherm models in most cases, and pseudo-first order or pseudo-second-order models are often more suitable for simulating adsorption kinetics than others. Adsorption mechanisms mainly include electrostatic interaction, ion exchange, complex formation, and redox reaction, and lots of modified carbon nanoadsorbents not only perform very well in laboratories but exhibit excellent applicability to real environmental water.

There are many factors affecting the adsorption performance of an adsorbent, such as its surface area, porosity, size, shape, composition, and functional groups. All of them can be utilized to optimize an adsorbent either by increasing the number and density of adsorption sites or by enhancing the interaction between adsorbents and adsorbates. Through analyzing the modification methods, functionalizing carbon nanomaterials with polymers, especially for those polymers containing electron-donating atoms, is becoming a very prevalent and promising way to improve their adsorption abilities due to the convenience in tailoring the molecular structures and properties of polymers. In order to facilitate the recovery of adsorbents after adsorption, making carbon nanomaterials magnetic is a more and more popular method owing to such characteristics of magnetic separation as easy to operate, economical, and environmentally friendly.

In short, many superior carbon-based nanoadsorbents have been developed that not only exhibit outstanding adsorption capacity and rate but allow for easy recovery. They have very promising uses in practice. However, there are still some issues to concern. Adsorbents with a high selectivity towards a particular metal need to be further developed. There are fewer reports that carbon-based nanoadsorbents can remove targeted metals in the presence of large quantities of other metals, especially when the coexisted metals are similar to the targeted ones. High-performance carbon-based nanoadsorbents that can achieve the removal of trace metals from aqueous solutions also need further development. In many cases, the concentrations of metals are extremely low and only a little higher than their maximum levels in drinking water. However, most adsorbents do not perform well for this concentration range. Moreover, the potential impacts of carbon-based nanomaterials on ecological and biological systems are rarely involved. The carbon-based nanoadsorbents can hardly be completely recovered or removed from the solution after adsorption. It is urgently needed to improve the ability of recovering or removing them by such methods as precipitation and filtration or to incorporate low toxic or non-toxic materials to reduce their toxicity, like biomass and polymers. In addition, there needs to investigate the technical and economical feasibilities of their mass production. Hopefully these issues can be paid attention to in the future work.