Introduction

Freshwater contamination by heavy metals can come from a variety of sources and be damaging to wildlife, alter landscapes, and impact human health even at trace levels [1]. Heavy metals do not degrade in nature, and when found in water, can flow into soil and contaminate many natural resources that both humans and animals depend on. Heavy metals contamination can result due to proximity to paved roads where storm water runoff moves contaminants from vehicles and road treatments to soils and freshwaters [2]. Mining activities threaten biodiversity and genetic variation of rivers [3] and result in carcinogenic metals such as chromium being distributed into nearby freshwater sources [4,5,6]. Additional sources of heavy metals contamination in freshwater include electroplating, metallurgy, chemical plants, and even agriculture [7]. Over time, due to heavy metals contamination of waters and soils, a significant accumulation of these metals in the human body can occur, and therefore, wastewater should be treated for heavy metals before water is returned to the environment [8, 9].

Adsorption techniques for heavy metals removal from freshwater are highly desirable due to economic feasibility [7]. Removal costs by various methods for various adsorbents were summarized by Adeleye et al. [10]. In their summary, heavy metals removal processes such as ultrafiltration, ion exchange, reverse osmosis, and use of activated carbon require a removal cost of approximately 0.021015 $/g to 0.228917 $/g, whereas removal by a nanomaterial, namely nanoscale-zero valent iron (nZVI), was estimated at 0.000753 $/g. This summary reflects a 99.7% reduction in cost. Recently, nanomaterials have gained attention as a promising alternative for heavy metals adsorption [7]. Nanomaterials are an attractive option due to the presence of emergent properties that result from the high surface-area-to-volume ratios of such materials. NPs are so small that they are influenced by quantum mechanical effects [11] and often contribute to unique properties as compared to bulk materials, including an altered melting point [12], fluorescence [13], electrical conductivity [14], and magnetic permeability [15]. In nanoscale particles, a greater percentage of the internal material is exposed on the particle surfaces owing to the extraordinarily high surface area to volume ratio. These properties contribute to the exceptional adsorption capacity and reactivity of NPs [7]. Magnetically susceptible NPs in particular have garnered great interest in metals extraction processes due to the capability for magnetic collection once the particle surface is coated in the metal to be retrieved. Silica-coated, magnetic composite materials have been highly studied for this purpose. Previously, researchers developed processes for coating magnetite NPs with silica (SiMNPs) to provide an adsorptive surface and demonstrated the application of these particles on an industrial scale [16, 17].

Despite these encouraging results, nanomaterial distribution into the environment is an area of concern [18]. Nano-particulate exposure has resulted in significant human health problems [19]. Due to the small size and reactivity, such materials are small enough to penetrate organs, tissues, or even cells; and the ease with which NPs can move through living systems is concerning. Researchers have found NPs from air pollution in the frontal cortex of autopsy brain samples [20], in the urine of healthy children [21], and even to the fetal side of the placenta [22]. This exposure has resulted in lower birth weights [23, 24], preterm birth [25, 26], and intrauterine growth restriction [27, 28]. Polystyrene NPs [29] and silver NPs [30] can also reach fetal blood circulation. Other examples of side effects from NM translocation in the body include carbon nanotube (CNT) inhalation, which has led to inflammation, pulmonary fibrosis, and genotoxicity. CNTs are also potential carcinogens [31]. Colloidal silver has been shown to cause antibiotic resistance [32], can result in oxidative stress, and can have toxic effects on marine species [33, 34]. Although vertebrate species have been exposed to naturally occurring NPs throughout time, these highly reactive particles could be more threatening than naturally occurring NPs because organisms have not had sufficient time to adapt to their unique properties [35, 36], and some studies suggest NPs persist in the environment longer than naturally occurring NPs [37].

These concerns make it imperative that any NPs used for heavy metals extraction processes be non-toxic should they end up in soils, drinking water, or food sources. One naturally occurring NM is hydroxyapatite (HA), which is the main mineral component of vertebrate bone and teeth, where it is present as nanoscopic crystals. Synthetic HA materials have been used extensively for decades in biomedical applications due to excellent biocompatibility and osteogenic capacity [38]. More recently, hydroxyapatite nanoparticles (HA-NPs) have been incorporated in oral care products to treat dentin hypersensitivity and promote enamel remineralization [39,40,41,42,43,44]. An article in Scientific Reports from 2019 shows that commercially available nanoscale hydroxyapatite particles were highly cytocompatible in vitro and did not possess any irritation potential [45]. This study concluded that the particles did not alter the normal behavior of the cells and therefore were safe to be used in oral care products.

Previously, synthesis of Fe-doped HA NPs (IDANPs) was accomplished [46,47,48,49]. The synthesis process required a simple wet chemical precipitation process, where ingredients were added to stirring water at room temperature. In this work, the synthesis process used for IDANPs was modified to produce magnetic NPs coated in HA for biocompatibility and functionalized with TiO2 to improve adsorptive capacity. The novel TiO2-functionalized, HA-coated magnetite NPs (TiHAMNPs) were characterized, evaluated for toxicity, and assessed for adsorptive capacity; and an adsorption isotherm model was evaluated. Adsorptive studies were carried out with Cu, a heavy metal contaminant that can cause irritation of nose, mouth, eyes, headache, stomachache, dizziness, and diarrhea [1]. In addition, adsorptive capacity of rare earth elements (REEs) by TiHAMNPs was also evaluated. REEs are critical to production of green energy and superior communication technologies and provide a significant challenge as separation of individual REEs is extremely challenging and expensive [50, 51].

Experimental section

Method 1: Synthesis of titanium dioxide functionalized hydroxyapatite magnetite nanoparticles (TiHAMNPs), hydroxyapatite-coated magnetite nanoparticles (HAMNPs) and silica-coated magnetite nanoparticles (SIMNPs).

TiHAMNPs were synthesized by wet chemical precipitation methods adapted from Andriolo et al. [46,47,48,49]. Precursors were added in the order listed:

  1. 1.

    200 mL deionized H2O (18 MΩ) was stirred by stir bar at room temperature (~ 25 °C) at 275 rpm.

  2. 2.

    0.260 g calcium hydroxide (Ca(OH)2) purchased from Sigma Aldrich was added to the stirred H2O until dissolved.

  3. 3.

    0.243 g magnetite NPs (Fe3O4) purchased from SkySpring Nanomaterials (98 + % purity, 20–30 nm diameter) was added to the Ca(OH)2 solution and let stir for 1 min.

  4. 4.

    0.263 g titanium dioxide (TiO2) nanopowder (anatase, 99.7% purity, < 25 nm diameter, purchased from Sigma Aldrich) was added to the stirred solution and let stir for 1 min.

  5. 5.

    0.408 g potassium phosphate (KH2PO4) purchased from Sigma Aldrich was dissolved separately in 50 mL deionized H2O (18 MΩ). The KH2PO4 solution was then added dropwise over 2 min to the stirred solution.

  6. 6.

    pH of the solution after steps 1–5 was measured at 10. A solution of 10% hydrochloric acid (HCl) in deionized (18 MΩ) H2O was used to reduce pH to 7.5.

  7. 7.

    The final solution was stirred for 1 week at 25 °C and 275 rpm.

After 1 week, TiHAMNPs were centrifuged at 2500 rpm for 10 min before the supernatant fluid above the pellet was removed and replaced with deionized H2O (18 MΩ). Concentration of the particles was determined by drying a sample in a weigh boat under constant ventilation and weighing. Final concentration of the TiHAMNPs was adjusted to around 3 mg/mL. The final solution was autoclaved for 40 min to sterilize.

For comparison, hydroxyapatite coated magnetite nanoparticles (HAMNPs) were used in these studies. The HAMNPs were also synthesized using a modified version of the wet chemical precipitation process adapted from Andriolo et al. [46,47,48,49]. The synthesis process for HAMNPs was identical with exception of step 4, in which, 0.263 g citric acid (C6H8O7) is added in place of TiO2.

Synthesis of SiMNPs has been described previously [16, 52]. The synthesis process for SiMNPs involved three primary methods each of which involves several steps. First, a silica coating method was employed to Fe3O4 NPs. The steps involved included use of degassed solvents to prevent oxidation and loss of magnetism and mixing under an N2 environment. The second method for SiMNP synthesis is a surface functionalization method where (3-chloropropyl)trimethoxysilane (CPTMS) and methyltrimehoxysilane (MTMS) were added to the NP mixture, and NPs were washed with tolulene (C7H8) and acetone (C3H6O). The third method used during synthesis of SiMNPs was a surface functionalization step where the NPs were suspended in a poly(allylamine)∙HCl solution and washed again with deionized H2O (18MΩ) and C3H6O.

Method 2: Field emission scanning electron microscopy (FESEM)

Samples for FESEM were prepared by drying under constant ventilation and adhered to an aluminum stub with carbon tape. Samples were gold coated for 1 min prior to observation. The FESEM used was a Tescan Mira 3.

Method 3: Energy-dispersive spectroscopy

Energy-dispersive X-ray spectroscopy (EDS) was carried out concurrently with FESEM analysis using an EDAX Octane Elect EDS detector.

Method 4: X-ray diffraction

NP samples were dried under constant ventilation before being processed by mortar and pestle and placed on a single crystal quartz sample holder prior to analysis. X-ray diffraction (XRD) was performed on a Rigaku Ultima IV X-ray diffractometer.

Method 5: Zeta potential

Zeta potential was measured by ZetaPlus Zeta Potential Analyzer from Brookhaven Instruments. The technique employed by the ZetaPlus is electrophoretic light scattering based on reference beam (modulated) optics and a dip-in (Uzgiris type) electrode system. During analysis, 1.5 mL of pre-vortexed NP sample was added to an optically clear cuvette. Readings per sample were collected in three runs, in which, ten cycles are averaged per run.

Method 6: Statistical analysis

Statistical analysis was performed using SigmaPlot software, Version 11.0. Data was processed by one-way analysis of variance (ANOVA). In accordance with software protocols, normality was verified and passed prior to all ANOVA analyses by the Shapiro–Wilk test.

Method 7: Metabolic and cytotoxicity assays

African green monkey kidney (Vero) cells were stored at -80 °C in a 5% dimethyl sulfoxide (DMSO) solution of minimal essential media supplemented with 10% fetal bovine serum and 1% penicillin–streptomycin antibiotic prior to use (MEM++). The Vero cell line originated from the kidney of a normal adult African green monkey taken in 1962 in Chiba, Japan [53, 54]. The Vero cell line is the most widely accepted continuous cell line by regulatory authorities and was used to produce polio and rabies virus vaccines [55]. For experiments, cells were grown and maintained in an incubator (37 °C, 5% CO2) in 75 cm2 flasks with MEM++. When confluent, Vero cells were lifted with trypsin (0.25%)-ethylenediaminetetraacetic acid (EDTA) and split.

Toxicity assays were performed on Vero cells with 24 h NP exposure times. For these tests, the NPs being tested were centrifuged, supernatant was removed, and the NPs were re-suspended in MEM++. For both assays, optical density (OD) was measured at 450 nm and background taken at 620 nm.

The BioVision LDH Cytotoxicity Colorimetric Assay Kit II was used to evaluate release of lactate dehydrogenase (LDH) from Vero cells according to the kit protocols. Cytotoxicity % was also calculated based on enclosed kit protocols according to the following equation:

$$\mathrm{Cytotoxicity }\left(\mathrm{\%}\right)=\left[\frac{\left(\mathrm{Test\ Sample}-\mathrm{Low\ Control}\right)}{(\mathrm{High\ Control}-\mathrm{Low\ Control})}\right]\times 100\%$$
(1)

The Biotium XTT (2,3-Bis 2-methoxy-4-nitro-5-sulfophenyl -2H-tetrazolium-5-carboxanilide inner salt) Cell Viability Assay Kit was used to determine cell metabolic activity. The enclosed protocols were followed with the exception of the final step where activated XTT solution is added to the test wells. The nanomaterials exhibit an optical signal that can be detected by the plate reader, and therefore, must be removed prior to the measurement. Just before the activated XTT solution is added to the wells, fluid from each well was removed and placed in a labeled microcentrifuge tube. The empty wells were each fed with 100 µL fresh MEM++ to temporarily maintain the cells in the wells. The microcentrifuge tubes were centrifuged at 21,000 rpm for 5 min to remove any remaining nanomaterial. Temporary MEM++ liquid was then removed from the wells, and the centrifuged liquid from the microcentrifuge tubes was aspirated above the NP pellet before being dispensed back into the original well. Protocols for the Biotium XTT kit were then followed as normal.

Method 8: Adsorption studies

In adsorption experiments, 50 mL of 0.1 g/L Cu or rare earth metals in deionized H2O (18 MΩ) solution was added to separate flasks that had been acid washed with a 5% nitric acid solution. The Cu source used for preparation was CuSO4∙5H2O. For the rare earth metals, the following chlorides were used as the source for metals: EuCl3, PrCl3, GdCl3, or LaCl3. After Cu or rare earth solution was added to the flasks, the following was also added to the flasks: (1) nothing (control), (2) 1.0 g/L TiHAMNPs, (3) 1.0 g/L SiMNPs, or (4) 1.0 g/L HAMNPs. All flasks were placed on a temperature-controlled shake table at 25 °C and 225 rpm. Samples were taken at 15, 30, 45, 60, 180, 360, 720, and 1440 min intervals in triplicate. All samples were centrifuged at 15,000 rpm to remove TiHAMNPs, SiMNPs, or HAMNPs. All samples were then diluted by 100X (to accommodate the detection range of the Hach spectrophotometer system) and analyzed for Cu concentration on a Hach DR3900 Laboratory Spectrophotometer for H2O analysis using Copper TNTplus® Vial Tests. The Hach DR3900 has an absorbance accuracy of 1% at 0.50–2.0 and wavelength accuracy of ± 1.5 nm with a wavelength range of 340–900 nm. Initially, tests were performed with 0.01 and 1.0 g/L Cu. However, both of those concentrations proved to be outside of the spectrophotometers limits of detection (too low and too high, respectively). Therefore, 0.1 g/L Cu was used in our final adsorption tests and was also used for Cu recovery testing in Method 11c for comparison purposes. For consistency, rare earth tests also were performed with TiHAMNPs at 1.0 g/L and the rare earth metal at 0.1 g/L. For determination of REE concentration, inductively coupled plasma-atomic emission spectroscopy (ICP-AES) was used. Prior to measurements, a linear slope calibration was determined between 0 and 10 ppm for all REEs used.

Method 9: Magnetic collection studies: examination of nanoparticle loss, and copper recovery

TiHAMNPs were tested in a simplified version of a magnetic collection pipeline system (MCPS) used for metals collection of flowing water (Fig. 1). In the MCPS used, liquid flowed through polyvinyl chloride (PVC) piping followed by a PVC inline pipe mixer. The pipe mixer was used to provide improved contact between metal ions and sorbent particles. Following the mixer, collection magnets were used to collect the NPs coated in adsorbed Cu. To examine magnetic NP collection, NP loss and NP based copper recovery, the pipeline system was tested in three distinct ways. Between tests, all components were disassembled, acid washed (7% HNO3), and rinsed several times with 18 MΩ deionized H2O before the next test was performed.

Fig. 1
figure 1

Depiction of the MCPS (not to scale. Real dimensions indicated). Outline A shows the experimental setup used for magnetic collection. In these tests, a funnel was filled and used to control flow into the rest of the system. Outline B shows the setup used for determining any loss of NPs that might occur in the MCPS. *Note that the magnet holder was removed for these tests. The funnel was also used for loss experiments. Outline C delineates the entire MCPS system used to determine Cu recovery. In the MCPS, NP and Cu solutions are poured into the system simultaneously by opening both valves at the same time. Flow then follows the path through the PVC inline mixer to improve contact between metal ions and sorbent particles

Method 9a: magnetic nanoparticle collection

To monitor propensity of the particles for magnetic collection, the pipeline system was simplified further by removal of the PVC piping and static mixer (Fig. 1A). In this way, loss prior to the magnet was not a factor in our study. Control over flow rate with a peristaltic pump was not used during our testing to eliminate any loss throughout the tubing. During testing, 0.25 L H2O containing 1.0 g/L NPs was poured into the pipeline at 103.41 cm/s (338 mL/s) directly before the collection magnets. Flow rate was maintained consistent by pouring the entire sample into a funnel just before the magnetic collection section of the system. After a flow test, the magnet holder was removed from the system, and remaining NPs were washed out, dried, and weighed to determine the amount of particle collected. Baseline values were collected by drying each particle type in a weigh boat and comparing to the appropriate starting weights of the nanomaterials used, accounting for anything that affected weight during preparation including scale variability, or loss from container.

Method 9b: examination of nanoparticle loss

Any particle retention in the PVC pipe or static mixer prior to the magnet ultimately results in an unaccounted loss that will reduce the apparent NP and metal recovery efficiencies. For these tests, the collection magnets were removed (Fig. 1B), and 0.25 L fluid containing 1.0 g/L NPs was poured into the system three times through a funnel. Following the test, NPs were collected, dried, and weighed to determine loss. Flow rate in the system was measured at 103.4 cm/s (338 mL/s) with an average residence time of 0.30 s. Triplicate controls were prepared in the same manner as samples run through the system; however, samples were not run through the system but instead dried after preparation. This process ensured the appropriate starting weights of the nanomaterials used and accounted for any variables that would affect weight during preparation, including scale variability or loss from the container.

Method 9c: copper recovery in complete MCPS system

In the final test, adsorption of the NPs in the complete MCPS was examined (Fig. 1C). For these tests, the 1 L of TiHAMNPs (1.0 g/L) in 18 MΩ deionized H2O and 1 L 18 MΩ deionized H2O containing dissolved Cu (0.1 g/L) were secured to the PVC system by a threaded attachment. The bottles of liquid were held upside down, and flow was regulated by valve. Once the bottles were secured to the system, both valves were opened, and flow was allowed through the assembled MCPS system. TiHAMNPs were tested three times, and three samples per individual test were collected for ICP-MS analysis in 50-mL tubes that had been acid washed with 7% HNO3 and rinsed with sterile, 18 MΩ deionized H2O.

Following collection, samples were acid digested for analysis on a Thermo Scientific iCAP Q inductively coupled plasma-mass spectrometer (ICP-MS) for trace Cu. Acid digestion consisted of adding 3 mL trace metal grade nitric acid to 2 mL collected sample, followed by 2 mL sterile, 18 MΩ deionized H2O. The sample/acid/H2O mix was held at 90 °C in a heat block for 4 h before the block was switched off and samples were left overnight to cool. The following day, samples were brought up to 50 mL with sterile, 18 MΩ deionized H2O. Copper standards, Cu controls, and blanks were used during analysis. The values collected on ICP-MS represented that which had not been adsorbed and collected in the MCPS by the metals-collection NPs; and therefore, removal % was calculated by subtracting these values from the amount of Cu run through the system. As in previous tests, all components were disassembled between all sample runs, acid washed (7% HNO3), and rinsed several times with 18 MΩ deionized H2O before the next test was performed.

Results

Results 1: Nanomaterials characterization

TiHAMNPs were characterized by FESEM and EDS to examine morphology and bulk composition. XRD data enabled identification of the TiHAMNP composition, and zeta potentials were also taken to establish surface charge of the particles. For comparison, FESEMs and zeta potential of SiMNPs were also taken. Results showed that TiHAMNPs were (Fig. 2A) approximately 20–50 nm in diameter (Fig. 2B), while SiMNPs (Fig. 2C) appeared slightly larger at 30–100 nm in diameter (Fig. 2D). Both TiHAMNPs and SiMNPs were uniform across the samples examined.

Fig. 2
figure 2

A Photograph of TiHAMNPs on a magnetic stir bar. TiHAMNPs were light brown in color and exhibited magnetic properties. B FESEM of TiHAMNPs. TiHAMNPs exhibited uniform morphologies and diameters ranging from 20 to 50 nm. C Photograph of SiMNPs on a magnetic stir bar. SiMNPs exhibit a darker color than TiHAMNPs. D FESEM of SiMNPs showing uniform morphology and similar diameters to that of TiHAMNPs, although some particles appeared larger (up to 100 nm) than the manufacturer reported 30 nm maximum diameter

EDS analysis of the FESEMs for TiHAMNPs indicated all assumed elements were present, while any unexpected elements were not observed. The EDS analysis used did not have high enough resolution to evaluate components on the nanoscale but was used to confirm the presence or absence of expected elements. It was assumed that TiHAMNPs were composed of magnetite, HA, and TiO2 as distinct compounds making up the particle structure. This assumption was based on reports of HA forming on the surface of magnetite NPs [56], and TiO2 dispersion on the surface of HA by simple stirring/heating methods [57]. XRD analysis of TiHAMNPs as compared to the exact precursors used during synthesis revealed excellent fit with a Rietveld Refinement (Rwp) value of 1.35% (Fig. 3). Peak values also showed a strong visual match between the three individual pre-cursors and final TiHAMNP structure, with little to no loss of significant peaks.

Fig. 3
figure 3

A XRD data taken directly from the instrument software. Rwp for all three precursors compared to TiHAMNPs was 1.35. B XRD data graphed from raw data taken from the instrument. Graph shows relative intensities of the precursor materials and TiHAMNPs. Collectively, XRD indicated the three individual precursors were present in the final composite TiHMNP structure

Zeta potential measurements of TiHAMNPs indicated a near neutral surface charge (Fig. 4). When comparing TiHAMNPs to precursors used during synthesis, TiHAMNP surface charge appears to be most influenced by TiO2. This structure is desirable, since the TiO2 component of TiHAMNPs is meant for metals collection at the surface of the particle. SiMNPs exhibited a positive surface charge, similar to that of TiHAMNPs. NPs with neutral surface charge may reduce loss during the metals extraction process due to decreased attraction to other charged particles in the environment. In addition, the neutral surface charge of TiHAMNPs may result in reduced toxicity in the environment. Positive NP surface charge has been associated with increased risk of toxicity [58], non-specific cellular uptake [59], and accumulation in organs [59]. Negative nanoparticle surface charge has been associated with formation of abnormal actin filament formation which reduced cellular adhesion, proliferation, and motility [60, 61]. The harmful effects of nanoparticle surface charge have been greatly reduced through neutralization of such charges by passive coatings such as polyethylene glycol coatings [62, 63] that act to neutralize [64] or greatly reduce nanoparticle surface charges [65].

Fig. 4
figure 4

Zeta potential measurements of TiHAMNPs along with precursors and SiMNPs. Results show that the final TiO2 nanopowder coating of the particles increased zeta potential of the TiHAMNPs significantly. TiHAMNPs exhibited a slightly more neutral surface charge than SiMNPs

Results 2: Biocompatibility of nanoparticles

In addition to exhibiting good metals adsorption properties [66], TiO2 is approved by the Food and Drug Administration (FDA) as a food colorant and for cosmetic applications. TiO2 has exhibited not only biocompatibility, but also capability for reducing toxicity of metals such as Cu [67], acting as an antimicrobial [68], and even reducing biofouling of H2O sources [69]. Fe3O4 has also been evaluated for toxicity and revealed no genotoxic effects, no effect on organs and blood biochemistry in a rabbit model, and even inhibition of MCF-7 breast cancer cell proliferation in a dose-dependent manner [70]. However, release of nanomaterials into the environment is of concern due to the high diversity and novel properties that arise from highly reactive high surface areas and extremely small sizes and/or dimensions [71]. Due to these characteristics, nanomaterials present a unique challenge for environmental scientists attempting to track these materials and predict toxicity or effect on ecosystems. For certain nanomaterials, toxicity has been observed to vary with additional variables such as particle number, surface area, body burden, and crystalline structure [72, 73]. Because tracking of nanomaterials in the environment is problematic, toxicity of nanomaterials should be examined in vitro. In this work, the process of examining biocompatibility of TiHAMNPs was studied with the use of both cytotoxicity and metabolic assays.

The cytotoxicity assay was used to evaluate release of LDH from cell membranes upon rupture or damage after a 24-h exposure period. In these tests, a positive control consisting of deliberately ruptured cells and a negative control consisting of cells grown normally in medium were used to calculate a percentage of cytotoxicity of the test material. Results from these tests showed no significant difference between particle type with 1.0 g/L SiMNPs exhibiting a cytotoxicity of 11.06%, while TiHAMNPs exhibited a cytotoxicity of 4.54%.

The metabolic assay used (XTT assay) allowed comparison of cell health between mammalian cells that had been exposed to NPs vs not been exposed to NPs (Fig. 5). In those tests, a statistically significant diminishment of cell health was observed when the mammalian cell line had been exposed to SiMNPs for 24 h as compared to both TiHAMNPs (Fig. 5a) and HAMNPs (Fig. 5b). No statistically significant difference in cell health was observed between the control and cells exposed to TiHAMNPs (p = 0.791) and HAMNPs (p = 0.837).

Fig. 5
figure 5

Results from XTT assays with Vero cells. A Metabolic activity was not significantly different between TiHAMNPs and the control (p = 0.791), while cells exposed to SiMNPs experienced significant diminishment in metabolic activity (p < 0.001). B Metabolic activity was not significantly different between HAMNPs and the control (p = 0.837), while cells exposed to SiMNPs again experienced a significant decline in metabolic activity (p = 0.001)

Results from these studies are encouraging and indicate high biocompatibility of TiHAMNPs. Importantly however, it should be noted that extended use of TiO2 NPs has raised concern due to nanotoxicity induced by oxidative stress under UV light, resulting in formation of reactive oxygen species [74]. Although we did not use these methods here, green nanotechnology methods have been and should be employed during synthesis of TiO2 as a precursor for TiHAMNP synthesis in the future. These processes are efficient, reduce the number of precursors required, and can employ the use of natural organisms such as plants, bacteria, or proteins for TiO2 NP synthesis [75].

Results 3: Adsorption testing for Cu and rare earth elements

In addition to being synthesized for improved biocompatibility, TiHAMNPs were synthesized for metals collection in H2O. Specifically, the TiO2 coating of the HA coated magnetite was added as a biocompatible metals-adsorption layer. In adsorption tests, NPs (1.0 g/L) were placed in copper solution containing 0.1 g/L Cu in deionized H2O (18 MΩ). A Cu-only flask was also used to obtain the baseline Cu value. The solutions were then secured to a covered, temperature-controlled shake table held at 25 °C and 225 rpm for 24 h. Over the 24-h period, samples were removed, NPs were centrifuged out of solution, and the remaining fluid was analyzed in triplicate for dissolved Cu concentration. Results showed TiHAMNPs reduced Cu in solution significantly as compared to both HAMNPs (p = 0.011) and SiMNPs (p < 0.001) (Fig. 6). After 24 h, TiHAMNPs had removed 72.32% Cu from the H2O, whereas HAMNPs had removed 67.19% Cu, and SiMNPs had removed 30.39%. These values indicate that for recovery of high-value metals, TiHAMNPs may be the preferred choice; whereas lower value metals may be collected by HAMNPs that are less labor intensive to synthesize.

Fig. 6
figure 6

Results of adsorption experiment. Results show that TiHAMNPs exhibited significantly higher adsorption than HAMNPs (p = 0.011) and significantly higher adsorption than SiMNPs (p < 0.001). After 24 h, TiHAMNPs had reduced the amount of Cu in solution by a 72.32%, while HAMNPs reduced Cu by 67.19%, and SiMNPs had reduced Cu by 30.39%

Using adsorption data from the 24-h experiment (Fig. 6), it is possible to determine the rate constant (\(k\)) for the adsorption process. This was accomplished by applying two different adsorption models, pseudo-first order (PFO) and pseudo-second order (PSO) adsorption, and determining which model best fit our data. In order to evaluate the PFO model, the adsorption density (\({q}_{t}\)) at each time interval was determined. We then determined a reasonable value for the magnitude of \({q}_{e}\) (equilibrium adsorption density) by plotting a logarithmic curve of \({q}_{t}\) vs. time (\(t\)) and used the logarithmic equation from the data to approximate \({q}_{e}\) at 48 h (2880 min). Using the approximated value of \({q}_{e}\) at 48 h (2880 min), the linearized form of the PFO model was plotted using \({\text{Eq}}.\)(2). When using the intercept of the linearized logarithmic equation of the PFO model to calculate the value of \({q}_{e}\), we found that \({q}_{e}<{q}_{t}\) for most data points, making this an illogical assumption. Furthermore, the magnitude of the extrapolated value of \({q}_{e}\) found using the logarithmic curve was significantly different than the value of \({q}_{e}\) determined using the linearized model, which indicated that the PFO model was a poor fit. The PFO model is often applicable over the initial 20–30 min of the adsorption process, followed by extrapolation of the experimental data to infinite time to obtain \({q}_{e}\), which is often not practical [4]. Following extrapolation, \({q}_{e}\) becomes an adjustable parameter that is determined by trial and error.

$${\text{log}}({q}_{e}-{q}_{t})={\text{log}}{q}_{e}-\frac{{k}_{1}}{2.303}t$$
(2)

The PSO model, in contrast, assumes that the rate-limiting step is a chemical sorption or chemisorption step and predicts behavior over the entire adsorption range. In this way, the adsorption rate is dependent on adsorption capacity rather than on concentration of adsorbate. Furthermore, when graphing the model using Eq. (3), there is no need to assume a value of the equilibrium adsorption capacity (\({q}_{e}\)). Instead, the linearized curve has an excellent fit that can be used to determine both the value of the equilibrium adsorption capacity (\({q}_{e}\)) and the value of and the rate constant \({k}_{2}\) (Fig. 7).

Fig. 7
figure 7

PSO plot describing adsorption density of TiHAMNPs. The \({q}_{e}\) was 87.92 mg/g, and \({k}_{2}\) was calculated to be 0.0003 g/(min*mg). For comparison, the adsorption density of SiMNPs was also calculated and showed that a higher concentration of SiMNPs (0.0010 g/(min*mg)) would be required to achieve adsorption similar to TiHAMNPs within the same time period

$$\frac{t}{{q}_{t}}=\frac{1}{{k}_{2}{{q}_{e}}^{2}}+\frac{1}{{q}_{e}}t$$
(3)

For comparison, the same model was used to characterize the adsorption density for SiMNPs. Results revealed a greater concentration of SiMNPs would be required to achieve similar adsorption within the same time period (Table 1).

Table 1 Parameters calculated (based on Fig. 7) for PSO used to determine adsorption density for TiHAMNPs and SiMNPs

In addition to Cu removal, preliminary tests were performed to examine removal of REEs. REEs are critical to production of high-tech materials, and industrial development in the areas of green energy and communications is becoming increasingly dependent on efficient extraction of these elements [50, 51]. REEs have similar physical and chemical properties [76] and are difficult to separate at feasible cost. REEs are divided into light and heavy groups. REEs with atomic numbers 57 to 63 are considered light REEs, while those with atomic numbers 64 to 71 are considered heavy REEs [77, 78]. Heavy REEs are less common and therefore typically considered more valuable than light REEs. In this adsorptive test, removal of Gd (heavy), Eu (heavy), Pr (light), and La (light) was monitored over the same 24-h period as used for Cu removal monitoring. Most promising, TiHAMNPs removed 97.22% Gd, 93.78% Eu, and 100% Pr, while leaving 23.47% La in solution after 60 min (Table 2). The difference in collection efficiencies indicates potential for selective separation. While preliminary studies are promising, we are currently evaluating TiHAMNPs more thoroughly for REE applications. For instance, based on the results obtained, REE adsorption by TiHAMNPs should be examined more closely between 45 and 180 min to narrow the time at which adsorption should be stopped so that efficient removal processes can be developed. In addition, more than one replicate per sample should also be used to provide statistical relevance of the data. Results also appear to show that TiHAMNPs approach 100% loading capacity after 12 h. Increased concentrations of TiHAMNPs may also be used to pull higher concentrations of REEs at earlier time points to improve efficiency of removal. For comparison, we have compiled the data to show adsorption of Cu vs. REEs by TiHAMNPs. Collectively, our TiHAMNP adsorption data indicate the adsorption of the four REEs studied may be more efficient than Cu adsorption by TiHAMNPs (Fig. 8). The data also shows a decrease in Gadolinium adsorption between 720 and 1440 min indicating desorption may be taking place. The discrepancy may also be due to variation that could be remedied through the addition of more replicates. Future work will include evaluation of the mechanisms involved in selective adsorption by TiHAMNPs, re-use of the particles, and removal of metals from complex solutions.

Table 2 Rare earth element removal from water by TiHAMNPs
Fig. 8
figure 8

Bar graph showing a comparison between removal of Cu vs. REEs by TiHAMNPs. The data shows that TiHAMNP adsorption of all REEs exceeds that of Cu from 15 to 1440 min. Preliminarily, this data indicates the efficiency of REE adsorption by TiHAMNPs may exceed that of TiHAMNP adsorption of Cu

Results 4: Adsorption isotherm

The methods used to determine the adsorption isotherm were the same as were used for Cu recovery and rare earth extraction tests. In this study, seven samples containing concentrations ranging from 1.0 to 10.0 g/L TiHAMNPs were added to flasks containing 2.0 g/L Cu in 20 mL deionized water (18 MΩ). In Fig. 9a, recovery from each concentration is graphed. The Cu removal achieved with 10.0 g/L of TiHAMNPs was 94.9%.

Fig. 9
figure 9

A Plot of Cu reduction as compared to concentration of TiHAMNPs. Results show near maximum concentration of Cu was removed (94.9%) in the presence of 10.0 g/L TiHAMNPs. B Linear Langmuir regression analysis for determining maximum adsorption capacity \({(\Gamma }_{m})\) for Cu on TiHAMNPs. The linearized Langmuir model yielded \({\Gamma }_{m}\) of 2.8 mmol/g

Results from the recovery values were used to establish a fit to either a Freundlich or Langmuir adsorption isotherm model [79]. For the TiHAMNPs, the best fit was established with a linear Langmuir-type adsorption isotherm, resulting in an \({R}^{2}\) value of 0.9997 (Fig. 9b). This best fit indicates that TiHAMNPs adsorb metals in a monolayer at the particle surface, with no subsequent stacking of adsorbed molecules. The Langmuir adsorption isotherm model is described by \({\text{Eq}}. (4)\):

$${q}_{e}=\frac{{KC*\Gamma }_{m}}{1+KC}$$
(4)

where qe is the equilibrium phase concentration of adsorbate adsorbed per mass of adsorbent (mg/g), K is the Langmuir adsorption equilibrium constant (L/mg), \(C\) is the equilibrium concentration of adsorbate (mg/L), and \({\Gamma }_{m}\) is the maximum adsorbent phase concentration of adsorbate when surface sites are saturated with adsorbate (mg/g). At 10.0 g/L TiHAMNPs, maximum removal of Cu was 94.9%. Using the linearized Langmuir model, \({\Gamma }_{m}\) was found to be 178.6 mg/g or 2.8 mmol/g.

Comparatively, this adsorption capacity falls on the high end of adsorptive materials including magnetic nanoparticles, natural materials, activated carbon, and non-zero valent iron (nZVI) used for Cu adsorption, among others. The two materials listed that exceed the capacity of TiHAMNPs in this list provided green manufacturing alternatives but also required additional processing steps to obtain natural components for synthesis [80]. To see a comparative list, see Table 3.

Table 3 Comparative list of adsorptive capacities for copper by various materials used for heavy metals collection

To ensure Cu precipitation did not occur during the experiment, pH was monitored. All samples exhibited a pH between 3.5 and 4.5 before and after the seven-day adsorption experiment. Using the Pourbaix diagram for Cu in an aqueous solution [98], these findings show that the Cu in each sample was within the phase stability region for Cu2+, indicating no Cu precipitation occurred. Furthermore, no Cu precipitates were visually observed in any of the samples or controls before or after the seven-day experiment.

In the linear Langmuir analysis, the following formula \(({\text{Eq}}. (5))\) describes the ratio of the adsorbate concentration (mg/L) to the adsorbed concentration (mg/g) at equilibrium:

$$\frac{{C}_{e}}{{q}_{e}}=\left(\frac{1}{{\Gamma }_{m}}\right){C}_{e}+\left(\frac{1}{{K}_{ads}{\Gamma }_{m}}\right)$$
(5)

Results 5: Magnetic collection and retrieval of nanoparticles

Magnetic collection in the system was monitored by running triplicate samples of the TiHAMNPs in deionized H2O (18 MΩ) at a volume of 0.25 L and concentration of 1.0 g/L through the magnetic portion of the MCPS only (configuration shown in Fig. 1A) at 103.4 cm/s. Results showed TiHAMNPs collection efficiencies in this configuration to be 89.0% (Std. Dev. 0.001). Magnetic collection was compared to that of SiMNPs at the same concentration, volume, flow rate, and configuration. SiMNP retention was 74.9% (Std. Dev. 0.026), revealing no statistical difference between the nanoparticle types (Fig. 10, p = 0.145).

Fig. 10
figure 10

In magnetic collection studies, TiHAMNPs (0.25 L at a concentration of 1.0 g/L) were run through the system (configuration shown in Fig. 1A) at a rate of 103.41 cm/s. Results showed a retention of 89.0%. Studies were repeated with SiMNPs and showed a retention of 74.9%. Results showed no statistically significant difference in magnetic collection between the particle types (p = 0.145)

Results 6: Nanoparticle loss in the magnetic collection pipeline system (MCPS)

Loss in the MCPS was monitored by removal of the magnetic portion of the MCPS and introducing flow by funnel into the system (Fig. 1B). After triplicate samples were run, results showed no statistically significant loss of TiHAMNPs (p = 0.100, Std. Dev. 0.023) or SiMNPs (p = 0.069, Std. Dev. 0.026) as compared to triplicate controls and no statistical difference between nanoparticle types (p = 0.145). Control samples were prepared and poured directly out for drying to account for any loss that occurred due to processing rather than running through the system. Therefore, no loss was used to calculate total Cu recovered.

Results 7: Total copper recovery from MCPS system

Total Cu recovery tests were performed by allowing flow of Cu solution (1 L at 0.1 g/L) and TiHAMNPs (1 L at 1.0 g/L) into the system simultaneously (Fig. 1C). The two solutions then combined just before the static mixer and metals collection portions of the MCPS. Samples of the depleted solution were collected leaving the system and therefore provided a direct measurement of the Cu not collected by the NPs. These collection values were subtracted from control values of Cu solutions prepared in the same manner and run through the system without the TiHAMNPs present. Each test was run three times per TiHAMNPs or SiMNPs, and three measurements per run were taken and averaged. Results showed a collection efficiency of 46.86% for TiHAMNPs through the system at a flow rate of 103.4 cm/s and an average residence time of 0.30 s. Under the same conditions, SiMNPs were run through the system and exhibited an collection efficiency of 52.75%, which was not statistically different from the collection efficiency of TiHAMNPs (p = 0.569). These results indicate that the flow system should be modified to better match the ideal conditions for adsorption by TiHAMNPs. Modifications to the system should include increased residence time as results showed that longer exposure (in adsorption isotherm or 24-h adsorption tests) resulted in significant adsorption by TiHAMNPs as compared to SiMNPs.

Conclusion

Metals removal from freshwater sources can improve water quality and allow recovery of metals that would otherwise be lost to the environment. Use of biocompatible nanomaterials for removal of these metals would provide not only a highly reactive adsorption method, but could be employed more generously and broadly should there be no worry of nanomaterial that may be lost to the environment during collection processes. In this work, a magnetic NP passivated with HA was synthesized with a TiO2 coating for metals adsorption and biocompatibility. The TiHAMNPs were synthesized by wet chemical precipitation methods and brought to physiological pH. The NPs were approximately 20–50 nm in diameter with a near neutral (slightly positive) surface charge. X-ray diffraction measurements reveal the TiHAMNP structure consisted of three individual components: magnetite, HA, and TiO2. Biocompatibility of TiHAMPs was evaluated by metabolic and LDH assay. Results of these studies revealed that TiHAMNPs had no statistically significant influence over metabolic activity of mammalian cells and < 5% cytotoxicity after exposure for 24 h. In adsorption tests, TiHAMNPs (1.0 g/L) adsorbed 73.67% Cu (0.1 g/L) when agitated for 24 h as compared to 33.51% removal by SiMNPs and 67.19% removal by HAMNPs. Adsorption of TiHAMNPs at equilibrium was also measured to evaluate the adsorption capacity of TiHAMNPs as well as the adsorption isotherm. Using a PSO model, the equilibrium rate constant for the adsorption of copper by TiHAMNP’s was determine to be 0.0003 g/(min*mg). For the TiHAMNPs, the best fit was established with a linear Langmuir-type adsorption isotherm, resulting in an R2 value of 0.9997. This best fit indicates that TiHAMNPs adsorb metals in a monolayer at the particle surface with no subsequent stacking of adsorbed molecules. Using 10.0 g/L TiHAMNPs, maximum removal of Cu was 94.9%. The linearized Langmuir model was used to determine the maximum adsorption capacity (\({\Gamma }_{m}\)) of 178.6 mg/g or 2.8 mmol/g. In a final adsorption study, TiHAMNP propensity for REE removal was also examined. Promising results show elemental selectivity by TiHAMNPs, indicating potential use of TiHAMNPs to separate La from other REE’s. Future studies will evaluate the extent of applications and REE types that TiHAMNPs could be used to separate and/or collect.

TiHAMNPs were also tested in a prototype MCPS meant for freshwater sources as a demonstration for the material being used in real-world applications. At 1.0 g/L (1.0 L) TiHAMNPs collected 46.86% Cu (1 L at 0.1 g/L). Flowrate through the MCPS was 103.4 cm/s, and residence time was 0.30 s. These results may be improved through modification of the system design. Magnetic collection and loss of NPs in the MCPS were also evaluated, revealing 89.0% retention of the particles. This value was not statistically different from commonly used silica-coated nanoparticles containing magnetic cores. TiHAMNPs were compared to SiMNPs and HAMNPs for toxicity and adsorptive properties.

Collectively, the results presented here report synthesis of a highly novel, adsorptive, and biocompatible NP that can be synthesized easily by adding precursors to stirring water at room temperature, while remaining a competitive option for metals removal as compared to the current technology.