Introduction

Dendrimers exhibit an extensive range of potential applications due to their extraordinary and adjustable physicochemical properties [1, 2]. Attaching functional groups, such as amine, carboxyl, and hydroxyl, as terminal groups of the dendrimers can result in a substantial increase in binding capacity to a variety of ions [3, 4]. These specific characteristics make dendrimers promising candidates as efficient adsorbents [5, 6]. Golikand et al. [7] synthesized triazine-based dendrimers with a poly(ethylene glycol) core to sorb Cu(II), Ni(II), and Zn(II). Anbia and Haqshenas [8] reported Pb(II) and Cu(II) adsorption to mesoporous carbon nitride functionalized with melamine-based dendrimer amine. Zhao et al. [9] synthesized polyamidoamine (PAMAM)-grafted cellulose nanofibril aerogels for Cr(VI) removal. Xu and Zhao [10] investigated PAMAM dendrimers of various generations and terminal functional groups for copper(II) binding. Diallo et al. [11] reported Cu(II) binding to PAMAM dendrimers with terminal amine groups. Cu(II) binding capacities of the PAMAM dendrimers were much greater than those of linear polymers with amine groups. However, separation from the liquid phase is difficult, and filtration, centrifugation, or gravitational separation is required.

Magnetic separation overcomes several limitations of filtration, centrifugation, and gravitational separation [12–16]. Magnetic nanoparticles (MNPs) can be modified using functionalized natural polymers or anthropogenic polymers to enhance surface reactivity for various applications. Dendrimer or polymer structures can be combined with MNPs to achieve facile magnetic separation and high adsorption [17]. Chou and Lien synthesized dendrimer-conjugated MNPs by combining dendrimers with MNPs for effective removal and recovery of Zn(II) from aqueous solutions [18]. Kim et al. reported magnetite-cored dendrimers (MDs) with TiO2 terminal groups for significant improvement in methyl orange decolorization [19]. The reusability of spent adsorbents is an important issue in practical applications [20, 21]. MDs as adsorbents can be regenerated by desorbing the ions bound to the terminals groups with acid treatment. However, the stability of the adsorbents can be hampered in repeated regeneration due to component loss and degradation under acidic conditions [22]. Especially in MDs, ferrous ions leached from the magnetic core can stimulate the production of reactive oxygen species that lead to a Fenton reaction and trigger cytotoxicity [23]. Wang and Lo synthesized mesoporous magnetic iron-oxide (γ-Fe2O3) as a Cr(VI) adsorbent and monitored the decreased Cr(VI) adsorption in five consecutive regenerations [12]. Singh et al. synthesized amine-functionalized Fe3O4 and performed repeated adsorption–desorption experiments [24], with the adsorption of Cr(III), Co(II), Ni(II), Cu(II), Cd(II), Pb(II), and As(III) gradually decreasing. Shahbazi et al. [25] synthesized functionalized SBA-15 mesoporous silica by melamine-based dendrimer amines and regenerated the adsorbent for the adsorption of Pb(II), Cu(II), and Cd(II). After four successive regeneration cycles, the adsorption decreased by 4.7 % for Pb(II), 6.3 % for Cu(II), and 5.7 % for Cd(II). Previous researches focused mainly on the experimental observation of adsorption efficiency during repeated regeneration cycles rather than the fundamental cause of the declined adsorption efficiency. Few tried to monitor the stability of the MDs in repeated adsorption–desorption cycles, and this needs to be done to determine the reasons for the declined adsorption.

MDs terminalized with amine groups were synthesized and characterized as efficient adsorbents that could be magnetically separated and recycled for reuse. The objectives of this research were to: (1) investigate the adsorption and desorption of heavy metal ions on synthesized MDs, (2) identify the effective period and optimal storage condition for the synthesized MDs, (3) evaluate the reusability of MDs in repeated adsorption–desorption processes, and (4) identify the deteriorated parts of MDs in the regeneration process.

Materials and methods

Materials

Ferrous sulfate heptahydrate (FeSO4·7H2O), 3-aminopropyltrimethoxysilane (APT, H2N(CH2)3Si(OCH3)3), ethylenediamine (EDA, NH2CH2CH2NH2), hydrochloric acid (HCl), methanol (CH4O), lead nitrate (Pb(NO3)2), and cadmium nitrate tetrahydrate (Cd(NO3)2·4H2O) were obtained from Sigma-Aldrich (USA). Ferric chloride hexahydrate (FeCl3·6H2O) and methyl acrylate (CH2CHCOOCH3) were purchased from Junsei (Japan). Ammonium hydroxide (NH3·H2O) was purchased from Daejung Chemical (Korea). All reagents were used as received without further purification.

MD synthesis

The MNPs in this research were synthesized by co-precipitation of Fe2+ and Fe3+ (Fe3+/Fe2+ ratio = 2:1) in the presence of an ammonia solution and treated under hydrothermal conditions [18, 26]. A Fe2+ and Fe3+ mixture solution was made by dissolving 2.7 g FeSO4·7H2O and 5.7 g FeCl3·6H2O into 100 mL of de-ionized (DI) water. Chemical precipitation was achieved by the addition of NH4OH solution dropwise to reach pH 10 at 25 °C under vigorous stirring with nitrogen gas purging through the solution during the reaction in order to avoid oxidation of MNPs. The black precipitates were heated at 80 °C for 30 min. The products were washed with DI water and methanol after the reaction.

Two grams of the synthesized MNPs was suspended in 200 mL of methanol. Then, 11.6 mL of APTs was added in drop and the mixture was stirred for 7 h at 60 °C. The products were washed with methanol several times using magnetic separation. In this research, the synthesized MNPs modified only with APT were identified as Generation 0 dendrimer (G0). Two grams of G0 was dispersed in 100 mL of methanol. Then, 20 mL of methyl acrylate was added in drop. The suspension was ultra-sonicated and stirred at room temperature for 7 h. The products were washed with methanol several times using magnetic separation and then dispersed in 20 mL of methanol. The suspension was stirred at room temperature for 3 h after 4 mL of EDA was added. After washing with methanol, the solid product was magnetically separated and the supernatant was decanted. The product was dried at 65 °C under vacuum.

Characterization

X-ray diffraction (XRD) patterns were obtained using a Ni-filtered CuKα source (λ = 1.5418740 Å, 40 kV, 100 mA) in the range of 2θ = 20°–80° (D/MAX RINT 2000, Rigaku). Magnetization curves were obtained from a vibrating sample magnetometer (VSM, Quantum Design, PPMS-14). The Fourier transform infrared (FT-IR) spectra were also recorded (UATR Two, PerkinElmer). Thermo-gravimetric analysis (TGA) measurements were performed under a nitrogen atmosphere from 25 to 800 °C at a heating rate of 5 °C/min (SDT Q600, TA Instruments). The specific surface area was analyzed with the Brunauer–Emmett–Teller (BET) model using N2 adsorption at −196 °C (3 Flex analyzer, Micromeritics). The pore size distribution was derived from the desorption isotherms on the basis of the Barrett–Joyner–Halenda (BJH) method. The zeta potential was also measured (Zetasizer Nano ZS90, Malvern). The size and morphology of the samples were characterized with a scanning electron microscope (SEM) (Nova nano SEM 450, FEI) and transmission electron microscope (TEM) (H-8100, Hitachi). Metal ion concentrations in the supernatant were determined with inductively coupled plasma optical emission spectrometry (ICP-OES) (Optima ICP-OES 8000, Perkin Elmer). The detection limits of lead, cadmium, and iron were 2.47, 0.55, and 0.97 µg/L, respectively. The organic carbon concentrations in the supernatant were measured with a total organic carbon (TOC) analyzer (TOC-LCPH, Shimadzu). The detection limit of the organic carbon concentration was 4 µg/L.

Determination of the storage condition and effective period

The produced MDs were stored in methanol or sealed in dry vials to determine a better storage method over an effective storage period. Batch lead adsorption experiments and XRD pattern measurements of the stored MDs were monitored once a week. For adsorption experiments, 10 mg of the MDs was added to a 40 mL lead solution and stirred for 6 h at room temperature. The experiments were performed at an initial pH of 5.0 and the initial lead ion concentration ranged from 10 to 100 mg/L.

Core and organic branch loss

Although low pHs are needed for better regeneration of the used MDs for desorbing adsorbed ions from MDs with proton displacement, organic branches and magnetite cores are also susceptible to decomposition at low pHs [22]. To determine the optimal HCl concentration for regeneration of the used MDs without critical decomposition, 100 mg of the MDs was added to 100 mL of 10−4, 10−3, 1, and 5 M HCl solutions. After continuous stirring for 3, 6, and 24 h, the supernatant was withdrawn to determine HCl concentration for repeated regeneration of MDs. At the determined HCl concentration to the desorption procedure, repeated regeneration experiments were performed. Iron and organic carbon leaching fraction from the MDs were analyzed during loaded metal desorption from the MDs.

MD regeneration

Regeneration of the adsorbent was performed with Cd2+ as an adsorbate. The extracting agent in this research was HCl. After 4 h of equilibration, the metal-adsorbed MDs were magnetically separated, added to 40 mL of the extraction solutions, and mixed for 3 h. Subsequently, the MDs were separated with magnetic separation and washed with DI water repeatedly to remove the residual solution for the next adsorption cycle.

Results and discussion

The crystalline structure and magnetic properties were determined via XRD and VSM. Seven typical peaks of magnetite were observed at both MNP and MD in Fig. 1a, confirming that formation of the dendrimer structure has no significant effect on the crystalline characteristics of the nanoparticles. The magnetization curves from the VSM measurement of MNP and MD with the applied field at 25 °C are presented in Fig. 1b. Hysteresis loops of both MNP and MD indicated typical super-paramagnetic behavior of no hysteresis and an “S” shape [27–29]. The saturation magnetization was 65.50 emu/g and 45.36 emu/g for MNP and MD, respectively. The dendrimer structure formation at MD led to a decrease in the magnetic strength [30, 31]. However, the magnetization of MD was sufficient for separation [32], and MD was efficiently separated within 1 min, as shown in Fig. 1c.

Figure 1
figure 1

a XRD patterns of MNP and MD, b magnetization curves of MNP and MD, and c MD dispersed in an aqueous solution and magnetic separation

Full size image

Figure 2 shows the morphology of MNP and MD. Synthesized MNPs have pseudo-cubic structures with length of 8.59 nm. After formation of the dendrimer structure, the average size of MDs was 11.22 nm. An approximately 2.63 nm size increase was observed after the formation of the dendrimer structure, but the shape remained the same, as shown in Figs. 2, 3a, b. The results from the elemental analysis are shown in Fig. 3c, d. The Si K peak was detected due to the APT grafted to the MNP surface after the formation of the dendrimer structure.

Figure 2
figure 2

TEM images of a MNP and b MD

Full size image
Figure 3
figure 3

SEM images of a MNP, b MD, and EDS spectra of c MNP, d MD

Full size image

Figure 4 shows the FT-IR spectra of MNP and MD. In general, the characteristic bands of magnetite are shown at around 600 and 400 cm−1, which correspond to the Fe–O bond vibration of iron cations at the tetrahedral and octahedral site, respectively [33, 34]. In Fig. 4, the adsorption bands were observed at 548.91 and 472 cm−1 for MNP and 545.97 and 488 cm−1 for MD, respectively, indicating the formation of magnetite [35–37]. The broad Si–O bands in the range of 900 to 1110 cm−1 were observed in MD, confirming the aminosilanization reaction on MNP [19, 38]. The bands assigned to C=O were due to the dendrimer structure at 1733.74 cm−1 [19]. In addition, the banding vibrations of –CH (at 2924.27 cm−1) and –NH2 stretching (at 3230.49 cm−1) are observed in Fig. 4 [39, 40]. Figure 5 shows the TGA results, N2 adsorption–desorption, and pore size distribution of MNP and MD. TGA analysis of the samples was used to determine the content of the organic branches. The curve was composed of four weight loss steps as shown in Fig. 5a. The first weight loss step below 200 °C can be explained by physically bonded solvent (water and methanol) desorption from the surface [22, 41]. The second weight loss at temperatures ranging from 200 to 340 °C was most likely related to de-hydroxylation (desorption of OH−) [42]. The drastic weight loss from 340 to 600 °C corresponds to the thermal decomposition of the 3-aminopropyl groups grafted to the surface. Lastly, the weight loss above 600 °C was associated with the breakout of the structured water [22, 43]. The weight loss observed at 800 °C was 6.52 and 11.15 % in MNP and MD, respectively. The weight loss in MD was much greater due to the organic branch on the surface of MNP in MD. Figure 5b shows the N2 adsorption–desorption isotherms and corresponding pore size distributions (inset) calculated with the Barrett–Joyner–Halenda method [44–46]. The specific surface area decreased from 103.23 m2/g before aminosilanization to 84.14 m2/g after dendrimer structure generation as the grafted aminosilane groups was anchored to the inner pore volumes [18, 47]. This decrease can be further supported by the decrease in pore size (from 12.97 to 12.04 nm) and pore volume (from 0.41 to 0.33 cm3/g) after aminosilanization.

Figure 4
figure 4

FT-IR spectra of MNP/MD

Full size image
Figure 5
figure 5

TGA results of a MNP/MD, N2 adsorption–desorption isotherms and pore size distributions (inset) of b MNP/MD

Full size image

No significant difference in maximal adsorption of lead was observed between two different storage conditions, as shown in Fig. 6. However, a new XRD peak corresponding to hematite appeared after some time. The original magnetic core consisted of magnetite, but part of the magnetite was oxidized to hematite, as shown in Fig. 7. Once the core of MDs begins to be oxidized, decrease in recovery rate occurs due to weakened magnetism [48]. The hematite peak was not detected until 56 days when the MDs were sealed in dry vials. Alternatively, the hematite peak was observed after 47 days storage in methanol. Therefore, storage of the synthesized MDs in a dry vial was more suitable than storage in methanol. The effective period of MD was experimentally determined to be 56 days. Thus, all MDs produced in this research were stored in the dry vial condition and used within 56 days.

Figure 6
figure 6

Change in maximal adsorption (Q m) of lead to MD in two different storage conditions (temperature: 25 °C; contact time: 6 h; lead: 100 mg/L; MD: 0.25 g/L)

Full size image
Figure 7
figure 7

XRD patterns of a MD, b MD stored in methanol (after 49 days), and c MD sealed dry in a vial (after 63 days)

Full size image

The effect of pH on the adsorption and Fe leaching fraction of MD is shown in Fig. 8a. Adsorption of Pb and Cd increased with a pH increase from 2.0 to 5.0. This can be explained by competition between H+ ions and metal cations. However, at a pH greater than 5.0, metal hydroxide precipitation was observed as Pb(OH)2 and Cd(OH)2. To ensure no precipitation, the other adsorption experiments in this research were performed at pH 5.0 [49, 50]. Fe leaching from the magnetite core of MDs at different pH is shown in Fig. 8a. The Fe leaching fraction decreased with a pH increase, and negligible Fe leaching was monitored at pH 4.0 and above.

Figure 8
figure 8

a Effect of pH on the adsorption and Fe leaching fraction of MD (temperature: 25 °C; contact time: 6 h; adsorbate concentration: 100 mg/L; MD: 0.25 g/L), b Effect of contact time on the adsorption to the MD (temperature: 25 °C; pH 5.0; adsorbate concentration: 100 mg/L; MD: 0.25 g/L)

Full size image

In order to estimate the equilibrium time and to determine the adsorption rate of heavy metals, the effect of contact time on adsorption was studied. The initial metal concentration was 100 mg/L, and the experiments were performed at pH 5 and room temperature under different equilibration times. In Fig. 8b, with an increase in the sorption time from 0 to 30 min, the adsorption of Pb(II) and Cd(II) dramatically rose. After 60 min of contact time, the q t increased slightly and the plot almost became a straight line after 240 min. Therefore, it is obvious that the equilibrium reached after 360 min of contact time. In order to further analyze the adsorption of heavy metals onto MDs, the kinetics of heavy metal ions adsorption to the MDs was fitted using pseudo-first-order kinetic and pseudo-second-order kinetic models:

$$ { \log }\left( {q_{\text{e}} \, - \,q_{\text{t}} } \right)\, = \,{ \log }q_{\text{e}} \, - \frac{{k_{1} }}{2.303}t $$
(1)

where q e and q t are the adsorbed concentration (mg/g) at equilibrium and at time t (min), respectively, and k 1 is pseudo-first-order rate constant (min−1).

$$ \frac{t}{{q_{\text{t}} }} \, = \,\frac{1}{{k_{2} q_{\text{e}}^{2} }}\, - \,\frac{1}{{q_{\text{e}} }}t, $$
(2)

where k 2 is the adsorption rate constant of second order kinetic models.

Table 1 shows the calculated kinetic parameters for lead and cadmium adsorption. The R 2 values of the pseudo-first-order kinetic model for the adsorption of Pb(II) and Cd(II) to the MDs were 0.78 and 0.66, respectively. Those of the pseudo-second-order kinetic model were 0.99 for Pb(II) and 0.99 for Cd(II). In both metal adsorption, q values from the experiments (q e, exp) and calculated from the pseudo-second-order model (q e, cal) was almost identical. It indicates that the pseudo-second-order kinetic model describes the adsorption better than pseudo-first-order kinetic model. The rate-determining step of the adsorption to the MDs is therefore chemical adsorption through valence forces by sharing or switching electrons between the MD and heavy metals [51–54].

Table 1 Kinetics of heavy metal adsorption to MD (temperature: 25 °C; pH 5.0; adsorbate concentration: 100 mg/L; MD: 0.25 g/L)
Full size table

Figure 9 shows the schematic diagram of metal ion adsorption to the MDs in this research. Once heavy metal ions (M2+) are reacted with MD, they form coordinate bonds by sharing a lone electron pair of the nitrogen atom in the terminal groups of MDs. Adsorption to the protonated MD is slower than adsorption to non-protonated MD, as the attraction between the nitrogen atom in –NH3 + and M2+ ions is weaker than the attraction between the nitrogen atom in –NH2 and M2+ ions [55]. Therefore, adsorption to the protonated MD can be regarded as the rate-determining step. Adsorbed heavy metal ions can be desorbed from MDs by reducing the pH using HCl. Figure 10 shows the change of zeta potentials as a function of pH. The zero point charge of the MNPs and MDs in this research was 6.6 and 8.4, respectively. The MNPs and MDs were protonated at pH <6.6 and pH <8.4, respectively, and the positive charge density on the surface decreased as the pH increased. Consistent with the pH effect on the adsorption, more NH2 groups were converted to NH3 + at a lower pH. The density of the NH2 groups on the outer surface of the adsorbent decreases and the adsorption decreases.

Figure 9
figure 9

Schematic diagram of metal ion adsorption to MDs in this research

Full size image
Figure 10
figure 10

Zeta potential change of MNP and MD as a function of pH

Full size image

Iron loss under various acidic conditions was tested with different desorption equilibration periods to determine the HCl concentration for repeated regeneration of MDs, as shown in Fig. 11. The iron loss was increased with increasing time and HCl concentration. In mild acidic conditions (below 10−4 M HCl), iron loss was minimal, irrespective of the equilibration time. Iron loss occurred 10−3 M HCl. Seventy-four percent of the iron was lost from MDs after 3 h in 1 M HCl. In 5 M HCl, nearly all of the iron in the MDs leached out. Therefore, 10−3 and 10−4 M HCl at 3 h were chosen to minimize the MD structure decomposition during desorption.

Figure 11
figure 11

Iron loss from MDs in various desorption periods at different HCl concentrations

Full size image

The results from equilibrium adsorption of Pb(II) and Cd(II) to MDs were fitted to Langmuir and Freundlich models [51, 56, 57].

The Langmuir isotherm model can be expressed as follows:

$$ \frac{{C_{\text{e}} }}{{q_{\text{e}} }}\, = \, \frac{1}{{K_{\text{L}} q_{\text{m}} }}\, + \,\frac{{C_{\text{e}} }}{{q_{\text{m}} }} $$
(3)

where q e is the amount of adsorbed metal ions by adsorbent at equilibrium (mg metal/g adsorbent), C e is the equilibrium metal concentration (mg/L), q max is the maximum metal uptake of the Langmuir isotherm (mg metal/g adsorbent) and K L (L/mg) is the Langmuir constant which is related to the energy of adsorption.

The Freundlich isotherm model can be presented by the following equation:

$$ { \log }q_{\text{e}} \, = \,n {\text{log}}C_{\text{e}} \, + \,{ \log }K_{\text{F}}, $$
(4)

where K F (mg/g) and n are the Freundlich isotherm parameters related to adsorption capacity and adsorption intensity, respectively.

Table 2 shows the model fitting results. The R 2 values for the Langmuir model fitting in single metal system were 0.94 and 0.99 for Pb(II) and Cd(II), respectively. The maximum adsorption of Pb(II) and Cd(II) from the Langmuir fitting in single metal system was 170.42 and 75.17 mg/g, respectively. In binary system, the maximum Pb(II) adsorption from the Langmuir fitting was higher than that of Cd(II) (Pb2+: 111.03 mg/g, Cd2+: 67.65 mg/g). Compared to the single metal system, the maximum adsorption of two metals in binary system decreased due to the competition for the limited adsorption sites in the MDs [58]. The R 2 values for the Langmuir model fitting in binary system were 0.92 and 0.97 for Pb(II) and Cd(II), respectively. The affinity order of adsorption on MD was Pb(II) > Cd(II), and this difference arose from the dissimilar chemical properties between the two metals. Lead has a higher Pauling electro-negativity than cadmium (Pb2+: 1.87, Cd2+: 1.69) [59]. Therefore, Pb(II) was subjected more to attraction to the lone electron pairs in donor atoms to form more stable complexes, thus showing higher affinity to the amine functional groups on the surface of the MDs [60]. In addition, Pb(II) has lower pKH (negative log of hydrolysis constant) than Cd(II) (Pb2+: 7.71, Cd2+: 10.1), confirming that Pb(II) is more effectively adsorbed through surface complexation or sorption reactions than Cd(II) [61]. Moreover, the different adsorption preferences onto the MDs for Pb(II) and Cd(II) can also be explained by (1) higher atomic weight (Pb2+: 207.2, Cd2+: 112.41) [59], (2) higher ionic radius (Pb2+: 1.21, Cd2+: 0.97) [62], and (3) larger Misono softness value (Pb2+: 3.58, Cd2+: 3.04) [63] of Pb(II).

Table 2 Equilibrium model parameters for heavy metal adsorption to MD
Full size table

Figure 12a, b shows the regeneration experimental results of MD for Cd(II) adsorption. After fifteen cycles of adsorption–desorption, the adsorption decreased to 55 % in both cases due to the loss in MD components and insufficient desorption. Compared to the first cycle, the desorption efficiency gradually declined by 17 % at pH 3.0 and 12 % at pH 4.0. During the regeneration process, iron was leached from the magnetic core of MD. More iron was leached from the magnetic core with the stronger acid. After fifteen cycles, the cumulative iron loss fraction was 0.11 with 10−3 M HCl and approximately 3.4 times greater than 0.33 with 10−4 M HCl. Similar to iron, organic carbon was leached from the organic component of MDs. The cumulative organic carbon loss fraction after fifteen cycles of regeneration was 0.06 with 10−3 M HCl and 0.05 with 10−4 M HCl. Figure 12c shows the remaining Fe and organic carbon fraction. The remaining iron content was highly dependent on the acidity while the remaining organic carbon content was not severely affected by the different HCl concentrations.

Figure 12
figure 12

Regeneration of Cd using a 10−3 M HCl, b 10−4 M HCl, c and remaining Fe and organic C fraction after regenerations using HCl solutions

Full size image

The higher concentration of HCl resulted in more deterioration of the MD structure, while a nearly negligible difference was observed in the desorption efficiency between 10−3 and 10−4 M HCl. Therefore, it is better to use a 10−4 M HCl solution for 3 h. To achieve better adsorbent reusability, the magnetic core of MDs needs to be more resistant to acidic conditions for enhanced stability and reusability.

Conclusions

In this research, MDs were successfully synthesized for heavy metal adsorption. The effective period of synthesized MDs was 56 days and storage in a dry vial was better than storage in methanol. After seven regeneration cycles, adsorption was maintained at approximately 80 % the original value, but it gradually decreased to 55 % after fifteen cycles. This decrease was caused by insufficient desorption of loaded heavy metal and adsorbent loss due to iron leaching from the MD cores and impairment of organic branches in the acidic condition during the desorption process. More iron loss was observed than organic branch loss during the desorption process. To enhance the stability in repeated MD use as environmental adsorbents, the magnetic core of MDs needs to be more resistant to acidic conditions.