Core–Shell NaHoF4@TiO2 NPs: A Labeling Method to Trace Engineered Nanomaterials of Ubiquitous Elements in the Environment

Understanding the fate and behavior of nanoparticles (NPs) in the natural environment is important to assess their potential risk. Single particle inductively coupled plasma mass spectrometry (spICP-MS) allows for the detection of NPs at extremely low concentrations, but the high natural background of the constituents of many of the most widely utilized nanoscale materials makes accurate quantification of engineered particles challenging. Chemical doping, with a less naturally abundant element, is one approach to address this; however, certain materials with high natural abundance, such as TiO2 NPs, are notoriously difficult to label and differentiate from natural NPs. Using the low abundance rare earth element Ho as a marker, Ho-bearing core -TiO2 shell (NaHoF4@TiO2) NPs were designed to enable the quantification of engineered TiO2 NPs in real environmental samples. The NaHoF4@TiO2 NPs were synthesized on a large scale (gram), at relatively low temperatures, using a sacrificial Al(OH)3 template that confines the hydrolysis of TiF4 within the space surrounding the NaHoF4 NPs. The resulting NPs consist of a 60 nm NaHoF4 core and a 5 nm anatase TiO2 shell, as determined by TEM, STEM-EDX mapping, and spICP-MS. The NPs exhibit excellent detectability by spICP-MS at extremely low concentrations (down to 1 × 10–3 ng/L) even in complex natural environments with high Ti background.


■ INTRODUCTION
The past few decades have witnessed significant advances in nanotechnology, from the controlled synthesis of nanomaterials to their applications in nanomedicine, 1,2 energy harvesting and storage, 3,4 and soil and water remediation. 5,6 Nanosafety and nanotoxicology have emerged as new research topics in response to increasing concerns regarding the potential adverse effects on humans and the environment exposed to nanomaterials intentionally, or inadvertently. 7,8 As one of the few nanoparticles (NPs) that have already been widely used in industry for decades, TiO 2 NPs have been heavily produced for a wide range of applications, such as pigments, sunscreens, cosmetics, medical implants, self-cleaning surfaces, photovoltaics, photocatalysts, antifogging surfaces, and wastewater treatment. 9,10 Because of this prevalence, it is crucial to understand the fate of engineered TiO 2 NPs in the environment to assess their risk and control pollution. Indeed, TiO 2 NPs have been predicted to have the highest environmental occurrence of all engineered NPs, and have been found in treated wastewater, sewage sludge, surface waters, sludgetreated soils, and sediments. 11 Environmental concentrations of Ti are strongly influenced by geogenic sources. In rivers with high concentrations of suspended matter (6.0−140.6 mg/L), the fraction of suspended Ti reached 62.3−88.6% (1.0−7.5 mg/g in terms of dry mass) with a strong correlation between the mass of suspended matter and the concentration of suspended Ti. 12 Xray fluorescence spectrometry determined the Ti content of soil samples from Ti mining sites to range from 0.47 to 2.80%, but the Ti was found to be of geogenic origin with no anthropogenic input. 5 Though some efforts have been devoted very recently to discrimination of engineered TiO 2 from natural Ti-bearing NPs by a multielement detection approach, 13 this high natural background makes quantification of released and bioaccumulated concentrations of engineered TiO 2 NPs extremely challenging in the absence of some functionalization of the NPs to facilitate their discrimination. Labeling approaches proposed to date, for a range of NP compositions, have included radiolabeling, 14 stable isotope enrichment, 15 chemical doping with a low-abundance element, 16 or barcoding with DNA fragments. 17,18 Each of these potential approaches has advantages and challenges, with cost and the scale at which the NPs can be produced being the major drawback of all. For TiO 2 NPs, stable isotope labeling with 47 Ti has been successfully applied for detection of the bioaccumulation of NPs in zebra mussels (Dreissena polymorpha) exposed for 1 h at environmental concentrations via water (7−120 μg/L of 47 TiO 2 NPs) and via their food (4−830 μg/L of 47 TiO 2 NPs mixed with 1 × 10 6 cells/mL of cyanobacteria). 15 Chemical doping is a promising approach to achieve a large amount (grams compared to milligrams for radiolabeling) of labeled NPs at an affordable cost. However, introducing new cations into the lattice of host materials may alter their physical and chemical properties, even if the concentration of the dopant is low enough that the crystal structure remains unchanged. 19 It was reported that strong structural inhomogeneity, and even a phase transition, can be induced when there is a large difference in size between the substituted cations and the host cations. 20,21 A mixed-phase material, rather than a homogeneous solid solution, could be obtained because of unsuccessful doping. An alternative approach is to make a core of the tracer element surrounded by a shell of the material of interest. This core−shell approach is preferred for toxicological and environmental fate studies, because the material that comes into contact with the environment or living organisms will be the surface material and should be a close analogue of the undoped material, assuming that factors such as NP density are not significantly altered and appropriate crystal phase/morphology can be obtained.
ICP-MS was recently adopted as a means to detect NPs at ultra low concentration, 22 thanks to the capacity for elementspecific analysis and the low detection limits (down to ng/L). However, ICP-MS fails to differentiate between engineered NPs composed of high abundance elements (e.g., Ti) and their natural counterparts. 23, 24 To ensure that only engineered TiO 2 NPs are identified in complex media, Ho core−TiO 2 shell (NaHoF 4 @TiO 2 ) NPs were synthesized with the lowabundance element Ho used as a chemical marker. The core−shell design was proposed to achieve a high dopant concentration for better detection while retaining the structural integrity of the NPs being investigated via the shell. NaYF 4 has been intensively investigated as a host for up-converting fluorescent materials with tunable particle sizes being demonstrated through the use of small NPs acting as nucleating seeds. 25 As an analogue of NaYF 4 , it was expected that the size of NaHoF 4 NPs could be similarly controlled, to achieve particles above the size limit for spICP-MS detection (i.e., >20 nm). 24 For this reason, NaHoF 4 was selected as the marker core, although it is dissimilar from the TiO 2 shell both in structure and in composition.
TiO 2 -coated NPs can be synthesized via a hydrothermal process, 26−29 or a sol−gel reaction on the NP surface, 30 and Caruso et al. even proposed a layer-by-layer method for coating TiO 2 onto polymer NPs. 31 Unfortunately, the hydrothermal conditions, or ultralow concentrations, make these approaches unfavorable for large-scale synthesis, a prerequisite for NPs for environmental studies. A sol−gel approach derived from the Stober method has also been reported recently for TiO 2 coating; 32 however, this method was less effective for the coating of TiO 2 onto dissimilar

ACS Applied Materials & Interfaces
Research Article nanostructures 33 and a post-thermal treatment was needed to achieve a crystalline TiO 2 layer.
In this work, we developed a templating method for the large-scale synthesis of NaHoF 4 @TiO 2 core−shell NPs, in which a sacrificial Al(OH) 3 layer was deposited onto the NaHoF 4 NP surface and then etched by HF or other fluorinated species via the hydrolysis of TiF 4 .

■ RESULT AND DISCUSSION
The synthesis of NaHoF 4 @TiO 2 NPs involved four steps (see the Experimental Section for further details). A typical procedure can be described briefly as follows. First, NaHoF 4 core NPs were obtained by thermolysis of Ho(CF 3 COO) 3 and NaCF 3 COO in a mixed solvent consisting of a high boiling point solvent, 1-octadecene, and a coordinative solvent, oleylamine ( Figure 1A), adapted from the approach published for NaYF 4 NPs and its analogues. 25 Al(OH) 3 was then deposited onto the NaHoF 4 NP surfaces to improve colloidal stability in polar solvents such as ethanol or water. 34 An aqueous solution of TiF 4 was introduced to the NaHoF 4 @ Al(OH) 3 NPs dispersion along with polyvinylpyrrolidone (PVP, M w = 360 000) before the ethanol/water solvent system was heated to 60°C with stirring and maintained at this temperature for 24 h. A subsquent addition of ammonia−water was followed by reflux at 100°C for 2 h. Finally, the white product was isolated by centrifugation, washed with ethanol and water, and kept in ultrapure water.
The X-ray powder diffraction (XRD) pattern shown in Figure 1B confirmed the presence of hexagonal phase NaHoF 4 and tetragonal phase (anatase) TiO 2 . Compared to the NaHoF 4 core NPs, however, an obvious broadening effect was observed for the TiO 2 phase, indicating a very small crystal size. Transmission electron microscopy (TEM) images in Figure 2 revealed a rough NP surface after coating with TiO 2 , as well as an increase in the mean particle size from 61.4 to 68.6 nm. It was also noted that the size distribution of the NPs broadened, reflected by the fact that the standard deviation increased to 16.8 nm from 6.4 nm. Energy-dispersive X-ray (EDX) spectroscopy was utilized to further confirm the coexsitence of Ti and Ho ( Figure S1, Supporting Information), as well as the core−shell structure of the NPs. Scanning transmission electron microscopy EDX (STEM-EDX) mapping allowed for the elemental distribution of the NPs to be determined. The elements from the NaHoF 4 core particle (F, Ho and Na) were observed to be encompassed by the Ti and O from the external TiO 2 shell ( Figure 3A−F). In addition, more Ti was detected at the edges of the NP than in the core. Elemental line profiling was also done using STEM-EDX across a single core−shell NP to map its cross-sectional distibution of elements (Na, Ho, F, Ti and O). As observed in the elemental mapping, higher counts for Ti and O were detected at the periphery of the NPs (approximately 5 nm in thickness), while stronger signals from F and Ho from the NaHoF 4 core were evident in the middle of the NP, clearly demonstrating that the NaHoF 4 NPs were coated with a layer of TiO 2 (Figure 3G−L). A thickness of ca. 5 nm for the TiO 2 shell layer was consistent with the 9 nm increase in average particle size observed by TEM ( Figure 2). Note that despite the use of an Al(OH) 3 template in this work, no Al was detected for the product of NaHoF 4 @TiO 2 by EDX, as is evident from Figure 3. NaHoF 4 NPs obtained in organic solvents could not be used directly for TiO 2 coating via a hydrolytic approach in ethanol, because they were inevitably covered by oleyamine and were thus dispersible only in nonpolar solvents; therefore, surface modification was required to make them dispersible in polar solvents. Additionally, there is a lack of interaction between the hydrophobic organic layer of the NaHoF 4 NPs and the TiO 2 crystallite, which is unfavorable for the heterogeneous nucleation of TiO 2 on the NaHoF 4 surface. 35,36 The deposition of an Al(OH) 3 layer not only removes the surface bound oleylamine, but also imposes a highly positive surface charge onto the NaHoF 4 NPs, 34 providing a stable colloid in ethanol with a concentration up to 2 mg/mL. Once the TiF 4 solution was added into the NPs suspension, an external Al(OH) 3 layer

ACS Applied Materials & Interfaces
Research Article moves the equilibrium of hydrolysis toward the formation of Ti(OH) 4 by reacting with the consequential HF or through anion exchange with TiF 6 2− , TiF 5 − or other fluorinated species. 37 As a result, the Al(OH) 3 layer was etched and a Ti gel formed around the NaHoF 4 NPs. Ammonia−water was subsequently introduced to catalyze the condensation of the Ti gel to form a TiO 2 layer on the NP surface. PVP was then used to protect the newly formed NaHoF 4 @TiO 2 core−shell NPs from potential aggregation. A decrease in the hydrodynamic size was observed by dynamic light scattering (DLS) after the condensation triggered by addition of ammonia−water ( Figure  S2), thus confirming the loss of the Al(OH) 3 layer.
Some of the NaHoF 4 @TiO 2 NPs showed a significantly different morphology and smaller size in comparison to the NaHoF 4 core NPs before coating (Figure S3), and particle size analysis by TEM also showed a broader size distribution after coating with TiO 2 (Figure 2). These results led to a hypothesis that NaHoF 4 NPs were not stable in the presence of H + or Al 3+ since these ions could break Ho−F bonds, resulting in the formation of H−F or Al−F bonds. 37 This is supported by their bond dissociation energies (Al−F 675 kJ/mol, H−F 569 kJ/ mol and Ho−F 540 kJ/mol). Increasing the temperature or polarity of the solvent would encourage the dissolution of NaHoF 4 . Indeed, NaHoF 4 NPs appeared to be less stable in dimethyl sulfoxide (DMSO) than in ethanol. Only TiO 2 NPs were observed by TEM and a weak NaHoF 4 signal was detected by XRD for the product when using DMSO instead of ethanol as the solvent during the shell formation stage and on increasing the synthesis temperature to 160°C ( Figures S4  and S5).
Because of the intrinsic mismatch of the NaHoF 4 and TiO 2 lattices, TiO 2 tends to grow on the NaHoF 4 NP surface via a granule mode to form a rough layer consisting of small particles, minizing the free energy of system. As shown in the

ACS Applied Materials & Interfaces
Research Article TEM images ( Figures 2C and 4), the NaHoF 4 @TiO 2 core− shell NPs exhibited a rough surface after coating with TiO 2 . High-resolution TEM images revealed that the outer TiO 2 layer is formed from small TiO 2 NPs with a size less than 5 nm. This is also reflected by the broadening of the diffraction peaks in the XRD pattern ( Figure 1B). The (011) facets of TiO 2 , with a d-spacing of 3.52 Å, were observed in HRTEM, and its corresponding diffraction peak at 25.3°appeared as the strongest peak in the XRD pattern, confirming the presence of anatase TiO 2 . The bandgap of NaHoF 4 @TiO 2 was determined as 3.7 eV ( Figure S6), slightly larger than the typical value of 3.20 eV for anatase TiO 2, although this likely reflects the influence from the ultrasmall particle size of TiO 2 .
The formation of a core−shell configuration is not only thermodynamically dependent on the interfacial energy between the core and shell materials but also sensitive to kinetic factors including the reaction rate, temperature and the amount of NPs serving as crystal seeds. Because of a lower critical free energy, heterogeneous nucleation requires a lower chemical potential than homogeneous nucleation. 35,38 In other words, a higher concentration (supersaturation) of the soluble crystallite is needed for homogeneous nucleation. Therefore, there is a concentration window to form the hybrid material, above the critical level for hetergeneous but below the level for homogeneous nucleation. As one specfic example of a hybrid material, NaHoF 4 @TiO 2 NPs are more likely to form if the concentration of TiO 2 crystallite falls within this concentration window during the condensation process. Excess ammonia− water would lead to a fast condensation process, and a high concentration of TiO 2 crystallite if there are not enough NaHoF 4 NP seeds to consume them from the solution phase. Pure TiO 2 NPs, instead of core−shell structures, would form as a result of an homogeneous nucleation. However, an insufficient amount of ammonia could not trigger the

ACS Applied Materials & Interfaces
Research Article condensation or ensure a reasonable time scale for the reaction. Our results indicated that NaHoF 4 @TiO 2 core− shell NPs can be obtained when the relative amounts of ammonia, ethanol and water are 2 mL:740 mL:60 mL, respectively. A mixture of NaHoF 4 NPs and ultrasmall TiO 2 particles was achieved if more than 2.5 mL of 35% ammonia− water was used ( Figure S7). A colorless gel covered product was obtained if no or less than 1 mL of ammonia−water was added.
The challenges for synthesizing core−shell NPs on the scale required for field environmental fate experiments were thus to avoid the heterogeneous nucleation at a high concentration and the fact that a hybrid structure is thermodynamically less favorable than the formation of two separate homogeneous NPs. Unlike reactive Ti precursors, such as alkoxides, an elevated temperature or a high pH value is required to speed up the hydrolysis of TiF 4 39 or to facilitate the crystallization of TiO 2. . Currently, TiO 2 -coated materials with improved crystallinity are typically synthesized by a hydrothermal approach with an extremely low concentration to avoid the formation of unwanted pure TiO 2 NPs at elevated temperatures 28,40 (Table 1). In this study, the Al(OH) 3 layer plays an important role when synthesizing NaHoF 4 @TiO 2 NPs on a large scale. In addition to providing the NPs with excellent colloidal stability, it also serves as a sacrificial template to confine the hydrolysis and condensation process of TiF 4 to within the space surrounding the NaHoF 4 NPs ( Figure 1A), thereby helping to preclude the formation of pure TiO 2 NPs. The rate of TiF 4 hydrolysis was accelerated at a relatively low temperature (60°C), without altering the pH value, because of the presence of the Al(OH) 3 layer preventing the condensation process. No product was isolated after the reaction was held at reflux for 24 h, in the absence of ammonia−water, and a gellike product was recovered by centrifugation at 6000g for 20 min after stirring at 60°C for 24 h following adjustment to a Figure 5. spICP-MS results of NaHoF 4 @TiO 2 NP dispersions in ultrapure (UP) water and river water. (A) Real-time Ho signal from NaHoF 4 @ TiO 2 suspension in river water; (B) real-time Ti signal from NaHoF 4 @TiO 2 suspension in river water; (C) real-time Ho signal from NaHoF 4 @ TiO 2 suspension in ultrapure water; (D) real-time Ti signal from NaHoF 4 @TiO 2 suspension in ultrapure water; (E) size distribution of NaHoF 4 component detected by spICP-MS; and (F) size distribution of TiO 2 component detected by spICP-MS. Stock suspensions of NPs were diluted 100 million times with ultrapure water and river water, respectively, from ca. 1.5 mg/mL to ca. 15 ng/L for spICP-MS measurements. River water was collected from the Worcester and Birmingham Canal, near the University of Birmingham, and was used without filtration.

ACS Applied Materials & Interfaces
Research Article pH of 4 ( Figure S8), confirming the mechanism and the useful window for optimal core−shell NP synthesis.
Large-scale synthesis of TiO 2 -coated NPs could be potentially achieved via a sol−gel approach, 32 despite the fact that a calcination process up to 500°C is required to obtain a crystallized TiO 2 phase, with the consequent risk that larger particle aggregates may form because of the sintering effect. This approach requires preliminary NP seeds to be colloidally stable in a basic environment, which makes it inapplicable for positively charged or polymer-coated NPs. Positively charged NPs would lose their stability with addition of the ammonia−water catalyst, whereas NPs stabilized by functional polymer could remain colloidally stable in a basic environment, but the polymer layer being tightly bonded to the surface may hinder the coating with TiO 2 ( Figure S9). PVP was used in this work, and it was expected to interact with NPs via weak van der Waals forces. Our results suggested that this losely bound polymer layer did not affect the TiO 2 coating. Because of the weak interaction between PVP polymer and the NPs, PVP can be readily removed by washing with water or replaced by other coatings (especially under highly alkaline conditions), allowing the surface to be made more representative of the TiO 2 particles used in food, commestics and other applications. 42,43 As expected, the NaHoF 4 @TiO 2 core−shell NPs indeed showed superior detectability on spICP-MS even in the presence of a high background of Ti up to the μg/L regime ( Figure 5). Our results indicated that the ionic Ti levels in river (canal) water is up to 300 ppb, 100 times higher than the amount present in ultrapure water. In addition, Ti-containing particles were also observed in blank river water using spICP-MS. However, no Ho-containing particles, and very limited ionic Ho (<0.5 ppb), were detected either in river water or ultrapure water ( Figure S10). To simulate the measurement conditions in an environmental study (e.g, quantification of release, accumulation, environmental transformations, or presence in effluent (or sludge) following treatment in a wastewater treatment plant, for example), we diluted the NaHoF 4 @TiO 2 NP suspension (stock concentration, ca. 1.5 mg/mL) 100 million times with ultrapure water or river (canal) water, yielding a particle concentration of ca. 20 471 NPs/mL or a mass concentration of Ho of 4.5 ng/L as measured by spICP-MS. With the low background of Ti and Ho in ultrapure water, spICP-MS exhibited an excellent capacity to detect both the Ti and Ho components of the NaHoF 4 @TiO 2 NPs, and mathematically provided an equivalent mean size (from the equivalent spherical volumes) of 68.0 and 54.8 nm for the NaHoF 4 core and TiO 2 shell, respectively ( Figure 5E, F) using the density of anatase TiO 2 bulk material (3.9 g/cm 3 ) and a calculated density for NaHoF 4 (3.99 g/cm 3 , see calculation in the Supporting Information). As the number of Ho-and Ti-containing particles detected were comparable, we can assume that the Ho and Ti components detected by spICP-MS come from the same core−shell NPs. This leads to an overall NP size of 78.2 nm for the NaHoF 4 @TiO 2 NPs ( Figure S11, see calculation in the Supporting Information), and subsequently a thickness of 5.1 nm for TiO 2 shell, which is very close to the value given by the size analysis of TEM images (4.2 nm) and by the element mapping by EDX (5 nm). Despite the slightly larger diameter achieved by spICP-MS than by TEM both for NaHoF 4 (68.0 nm vs 61.4 nm) and for NaHoF 4 @TiO 2 (78.2 nm vs 69.7 nm), the results obtained by these two methods are convergent, if taking into account the fact that an underestimated value could be given by size analysis on TEM due to the low contrast (electron density) of TiO 2 and the nonspherical shape of particles, while an overestimated size could be yielded by spICP-MS if the actual density of the NaHoF 4 core is higher than the calculated value.
Not unexpectedly, spICP-MS was no longer able to detect the Ti component of the NaHoF 4 @TiO 2 NPs in river water, because of the much higher abundance of background Ti than in ultrapure water ( Figure 5A and B); however, the Ho component of NaHoF 4 @TiO 2 NPs was easily detectable in the river water. Real signal intensity did not show much difference in river water or in ultrapure water, in terms of the frequency (particle number) and the intensity of the Ho peak (particle size) ( Figure 5C, D). More importantly, the size of the Ho component (in the form of the NaHoF 4 core NP) detected under the different conditions (ultrapure water and river water) are the same, 68.0 nm ( Figure 5E). A similar result was obtained across a wide NaHoF 4 @TiO 2 NP concentration range both in ultrapure water and in river water (data now shown). In addition to the low abundance of the marker element (Ho), the long-term physical and chemical stability of the marker NP (NaHoF 4 ) is also crucial, because the leaching of Ho would result in an underestimated value for TiO 2 in the environmental samples. Only negligible ionic Ho (0.015 mg/ mL) was detected in the suspension of NaHoF 4 @TiO 2 NPs (ca 60−100 mg/mL), even after storage for over 14 months, which could be partially attributed to the core−shell structure, wherein the TiO 2 shell provides a barrier to the release of Ho.
These results demonstrated that spICP-MS is a sensitive and reliable technique to monitor Ti-containing NPs in complex environmental samples using Ho as a marker. This strategy could be extended to spICP-MS detection of NPs containing other nanomaterials containing elements of high natural abundance such as iron or zinc.

■ CONCLUSION
A novel approach to gram-scale synthesis of NaHoF 4 @TiO 2 core−shell NPs was achieved, as a new strategy to detect NPs containing elements of high natural abundance such as Ti in complex environmental samples by spICP-MS. The deposition of an Al(OH) 3 layer around the Ho core was crucial for the synthesis of NaHoF 4 @TiO 2 NPs, not only because of the excellent colloidal stability it provided in ethanol or water, but also because of the hydrophilic surface necessary for the effective TiO 2 deposition and coating. More importantly, the Al(OH) 3 layer acted as a sacrificial template which facilitated the separation of the hydrolysis and condensation of TiF 4 and confined these processes to the immediate vicinity of the NaHoF 4 NP surface, allowing for the deposition of the TiO 2 shell onto the NaHoF 4 NP surface. Even when using these approaches, the TiO 2 grew in a particular mode to form a noncontinuous phase on the NaHoF 4 NPs, which minimized the surface energy at the interface because of their mismatching lattice energies, resulting in NaHoF 4 @TiO 2 NPs with rough surfaces. Although they were dissimilar in structure, the affinity of Ti to F is very high such that strong chemical interaction between TiO 2 and NaHoF 4 was expected and observed.
Due to the Al(OH) 3 layer, this approach allowed for the large scale synthesis of NaHoF 4 @TiO 2 NPs, enabling their application in environmental studies of TiO 2 NP fate and behavior. The core−shell structure was confirmed by high-

ACS Applied Materials & Interfaces
Research Article resolution TEM and STEM-EDX mapping, as well as by spICP-MS. We demonstrated that these core−shell NPs remain detectable by spICP-MS in the presence of a high background of Ti despite the NPs being present at an extremely low concentration. The introduction of a low abundance element (Ho) as a tracer, without altering the structure of the particles, provided an effective solution for the detection of engineered TiO 2 NPs in the environment. This methodology will benefit research in nanotoxicology and ecotoxicology, and could also be a potential solution to the challenges of detecting other engineered NPs of high abundance elements such as Zn and Fe in the environment.
■ EXPERIMENTAL SECTION Materials. All chemicals and solvents were purchased from Sigma Aldrich and used without further purification. Ultrapure water (18.2 MΩ cm at 25°C) was obtained from a MiliQ purification system. River water was collected from the Worcester and Birmingham Canal near the University of Birmingham (UoB), and used immediately after collection without filtration.
Characterization and Synthesis of the NPs. Unless stated otherwise, all characterization was performed at UoB. X-ray powder diffraction data was collected on a Bruker D8 advance diffractometer with a copper target (λ = 1.5406 Å, 40 kV, 30 mA). All samples were prepared by drying 0.5 mL of aqueous solution onto an Si zero background holder in air. The parameters for a typical experiment are as follows: starting angle (2θ), 20°; stop angle, 80; step size, 0.02026°; time/step, 0.8 s; no. of scans, 3030; time of scanning, 42 min and 25 s. Hydrodynamic size and zeta potential were measured on a Zetasizer Nano ZS ZEN 3600 from Malvern. Single-particle ICP-MS (spICP-MS) data were obtained on a PerkinElmer NexION 350X. Transmission electron micropscopy (TEM), energy-dispersive X-ray (EDX) spectroscopy and elemental mapping were carried out at the Nanoscale and Microscale Research Centre, University of Nottingham on a JEOL2100F transmission electron microscope operating at 200 kV (field emission electron gun source, information limit 0.19 nm). EDX mapping was perfomed using an Oxford instruments XMax 80 T silicon drift detector with INCA Energy 250 Microanalysis system in conjunction with the JEOL digital STEM system.
Step 1: Synthesis of NaHoF 4 NPs. Ho(CF 3 COO) 3 was obtained by dissolving Ho 2 O 3 in trifluoroacetic acid (ca. 30% w/w) at 90°C followed by removal of solvent on a rotary evaporator to obtain a pink powder. Ho(CF 3 COO) 3 (8 mmol, 4.0 g) and NaCF 3 COO (11.8 mmol, 1.6 g) were dissolved in a 250 mL round-bottom flask containing oleayamine (40 mL) and 1-octadecene (40 mL) before being heated to 120°C for 30 min in vacuo. After flushing with N 2 three times, 70 mL of the solution was removed by syringe and the rest of the solution in the flask was put on a preheated metal bath with stirring at 310°C under an atmosphere of N 2. The 70 mL aliquot was slowly injected back into the system over a 30 min period with continuous stirring under N 2 . Once all of the solution had been transferred, the temperature was lowered to 300°C and the reaction system held at this temperature for 1 h before being cooled to room temperature. NaHoF 4 NPs precipitated from the solution by the addition of ethanol (200 mL), and were isolated by centrifugation prior to their redispersion in hexane (300 mL).
Step 2: Synthesis of NaHoF 4 @Al(OH) 3 NPs. To the NaHoF 4 NP dispersion in hexane (300 mL), was added oleylamine (2 mL) with stirring at room temperature. The dispersion remained clear afterthe addition of a diethyl ether solution containing AlCl 3 (5 mL, 1 g/mL). After stirring for 10 min, water (5 mL) was added dropwise, and the clear dispersion became more and more opaque to form a white cloudy but stable colloid. NPs were precipitated out by addition of acetone (300 mL), and collected by centrifugation.
Step 3: Synthesis of NaHoF 4 @TiO 2 NPs. The NaHoF 4 @ Al(OH) 3 NPs were dispersed in ethanol (740 mL) in the presence of PVP (4 g, Mw = 360 000). An aqueous solution of TiF 4 (60 mL, 25 mmol) was quickly added into the ethanol dispersion of NaHoF 4 @ Al(OH) 3 NPs under stirring at 60°C, resulting in a gradual color change from pink to a yellow-green within 5 min. After stirring at 60°C overnight, the dispersion became slighly milky, indicating the formation of the Ti gel. The system was brought to reflux by heating to 100°C. After the quick addition of ammonia−water (2 mL, 35% w/w), the solution became cloudy. A white product was achieved by centrifugation at 6000g for 30 min, which was subsequently washed with ethanol and water, and finally stored in water. The yield of NaHoF 4 @TiO 2 NPs was calculated to be approximately 70% in terms of Ti.
spICP-MS Analysis of NaHoF 4 @TiO 2 NPs. The NPs were diluted 100 million-fold using Milli-Q or canal water to obtain the final concentration for analysis. The dilution was chosen from a preliminary dilution test with dilution ranging from 10 000 to 10 000 000 000, from which it was found that 100 million-fold dilution brought the NP concentration to within 5000−200 000 particles mL −1 (this being the desired range for spICP-MS analysis). The elements within the NPs were analyzed sequentially, and the river (canal) water samples were run last in order to avoid any carryover effects that may occur due to the high background of Ti in the river water. The instrument was calibrated using PerkinElmer Setup Solution. Ti and Ho were calibrated using ionic solutions obtained from Aristar and PerkinElmer, prepared as a dilution series to form a calibration curve. Finally, the transport effiency was calculated using 20 and 40 nm gold NPs obtained from Nanocomposix and gold ionic solution from Aristar.

Author Contributions
The manuscript was written through contributions of all authors. X.C. designed and carried out the synthesis and characterization of materials, analysized the data, and prepared the draft. B.F. and X.C. recorded and analyzed the spICP-MS results. D.Z. and X.C. analyzed the particle size and the size distribution. R.W.L., X.C., and A.N.K. carried out imaging and elemental analysis on TEM. I.L. and X.C. designed the study, I.L. and E.V.J. supervised the project, X.C. and I.L. wrote the first draft of the manuscript. All authors have given approval to the final version of the manuscript.

Funding
This work was supported by the European Union via the Horizon 2020 project NanoFASE (Grant 646002) and a Natural Environment Resaerch Council CASE PhD studentship with PerkinElmer as the CASE partner (BF).

Notes
The authors declare no competing financial interest.

■ ACKNOWLEDGMENTS
Intensive discussions with Ralf Keigi and Alexander Gogos (EAWAG) in terms of the NP specifications required for

ACS Applied Materials & Interfaces
Research Article quantification of TiO 2 NPs recovered from wastewater treatment plants (for example) within the NanoFASE project are gratefully acknowledged. The authors acknowledge the Nanoscale and Microscale Research Centre, University of Nottingham, for access to TEM facilities.