Distinguishing Engineered TiO₂ Nanomaterials from Natural Ti Nanomaterials in Soil Using spICP-TOFMS and Machine...

Area of Science:

• Environmental Science
• Materials Science
• Analytical Chemistry

Background:

• Distinguishing engineered nanomaterials (ENMs) from natural nanomaterials (NNMs) in complex matrices like soil is challenging due to elemental similarities.
• Titanium dioxide (TiO2) is a common ENM, and its presence in soil requires accurate identification methods.

Purpose of the Study:

• To develop and validate a machine learning approach for identifying TiO2 ENMs in soil.
• To differentiate TiO2 ENMs from naturally occurring titanium-based NNMs using spICP-TOFMS data.

Main Methods:

• Utilized single-particle inductively coupled plasma time-of-flight mass spectrometry (spICP-TOFMS) to analyze elemental fingerprints and mass distributions of TiO2 ENMs and Ti-NNMs.
• Developed and trained machine learning models, including a logistic regression (LR) model, to classify ENMs and NNMs based on spICP-TOFMS data.
• Investigated elemental associations and Ti-mass distributions for both synthesized ENMs and NNMs extracted from various soil types.

Main Results:

• Machine learning models showed best performance when differentiating based on Ti-mass distribution.
• A trained LR model could accurately classify 100 nm TiO2 ENMs at concentrations of 150 mg kg-1 or higher.
• Synthesized TiO2 ENMs were largely unassociated with other elements, while a significant fraction of Ti-NNMs showed no measurable associated elements.

Conclusions:

• Machine learning models, particularly those using Ti-mass distribution, can effectively identify TiO2 ENMs in soil.
• While promising, the inherent variability in chemical fingerprints of both ENMs and NNMs introduces uncertainty in identification and quantification.
• This approach offers a viable method for confirming the presence of TiO2 ENMs in most soil environments.