Application of Isotopically Labeled Engineered Nanomaterials for Detection and Quantification in Soils via Single-Particle Inductively Coupled Plasma Time-of-Flight Mass Spectrometry

Abstract

Finding and quantifying engineered nanomaterials (ENMs) in soil are challenging because of the abundance of natural nanomaterials (NNMs) with the same elemental composition, for example, TiO 2 . Isotopically enriched ENMs may be distinguished from NNMs with the same elemental composition using single-particle inductively coupled plasma timeof-flight mass spectrometry (spICP-TOF-MS) to measure multiple isotopes simultaneously within each ENM and NNM in soil, but the minimum isotope enrichment needed for detection of ENMs in soil is not known. Here, we determined the isotope enrichment needed for 47 Ti-enriched TiO 2 ENMs to be detectable in soil and assessed the effects of weathering on those requirements for less soluble TiO 2 and more soluble CuO ENMs. The isotope-enriched ENMs were dosed into two different soils and were extracted and measured by spICP-TOF-MS after 1, 7, and 30 days. Isotopeenriched ENMs were recovered and detected for all three time points. The 47 Ti-enriched TiO 2 ENMs were detectable in Lufa 2.2 soil at a nominal dosed concentration of 10 mg-TiO 2 kg −1 which is an environmentally relevant concentration in biosolid-amended soils. For distinguishing an ∼70 nm diameter TiO 2 ENM from TiO 2 NNMs in Lufa 2.2 soil, an ∼10 wt % 47 Ti isotope-enrichment was required, and this enrichment requirement increases as the particle size decreases. This study is the first to evaluate the tracking ability of isotope-enriched ENMs at an individual particle level in soil and provides guidance on the isotope enrichment requirements for quantification of ENMs made from Earth-abundant elements in soils.

Characterization and Quantification of Natural and Anthropogenic Titanium-Containing Particles Using Single-Particle ICP-TOFMS

“…Minimization and control of false-positive Ti-eng classifications are especially important for the analysis of environmental samples, which may have low PNCs of Ti-eng particles and can have variable PNCs of Ti-nat particles. Unlike previous approaches, , our Ti-eng particle classification does not depend on Ti particle size as a distinguishing characteristic. Instead, classification of Ti-eng particles is based on the presence or absence of secondary elements and the predicted mass fraction of these elements in Ti-nat particles.…”

“…Multielemental fingerprinting with spICP-TOFMS was first demonstrated by Praetorius et al to distinguish cerium-containing natural particles from CeO 2 engineered particles . Since then, multielement fingerprinting by spICP-TOFMS has emerged as a preferred approach to classify NPs and μPs at the single particle level, ,− although single-particle XRF studies have also been performed . For the characterization of the anthropogenic fraction of Ti-eng particles in environmental samples, most of the existing ICP-MS methods rely on bulk analysis.…”

“…For the characterization of the anthropogenic fraction of Ti-eng particles in environmental samples, most of the existing ICP-MS methods rely on bulk analysis. In these methods, bulk concentration ratios of, for example, Ti:Nb, Ti:Al, Ti:V, or Ti:Y are used to indirectly quantify the input of Ti-eng particles, which would elevate each of these ratios with respect to their crustal abundance ratios. ,,, Machine learning (ML) models have also been developed to classify and quantify individual Ti-eng and Ti-nat particles in soil samples based on Ti mass amounts and elemental fingerprints. , These ML approaches require well-characterized training sets from particles whose compositions and size distributions match those expected to be found in environmental samples. Researchers found that these ML models are unable to distinguish some Ti-nat and Ti-eng particle types when Ti-containing NPs have overlapping Ti-mass distributions.…”

“…16,19,38,55 Machine learning (ML) models have also been developed to classify and quantify individual Ti-eng and Ti-nat particles in soil samples based on Ti mass amounts and elemental fingerprints. 49,56 These ML approaches require wellcharacterized training sets from particles whose compositions and size distributions match those expected to be found in environmental samples. Researchers found that these ML models are unable to distinguish some Ti-nat and Ti-eng particle types when Ti-containing NPs have overlapping Timass distributions.…”

Machine learning analysis to classify nanoparticles from noisy spICP-TOFMS data

“…25 Previous studies have also used a variety of methodologies to characterize NP events for source apportionment. 1,18,19,21,23,25,27,[32][33][34][35][36][37][38] For spICP-TOFMS analysis, particle classication has been performed with supervised and unsupervised machine learning, among other methods, because of the potential for automated labelling and classication, thus, reducing the analysis time. 21 Examples of supervised learning methods used for NP and mP classication from spICP-TOFMS datasets include gradient boosted classiers (GBC), 32 light-GBC, 21 k-nearest neighbor embedding (KNN), 35 and binomial logistic regression (LR).…”