# Strategic Brazilian Minerals Applied in the Photocatalytic Ozonation of Rhodamine B Using a Green Lithium Niobate Nanocatalyst Supported on Silica: Kinetic, Thermodynamic, Mechanism, Machine Learning, and Ecotoxicity Study

## TL;DR

This study creates an eco-friendly nanocatalyst from rice husk and lemon to efficiently break down a harmful dye using light and ozone.

## Contribution

A green nanocatalyst made from agro-industrial waste and Brazilian minerals for efficient dye degradation via photocatalytic ozonation.

## Key findings

- The SiO2/LiNbO3 nanocatalyst achieved over 97% Rhodamine B degradation in under 60 minutes.
- Superoxide radicals and valence holes were identified as the main degradation agents.
- The KNN machine learning model accurately predicted degradation intermediates.

## Abstract

This study presents
the synthesis and use of a novel heterogeneous
lithium niobate nanocatalyst supported on silica (SiO2/LiNbO3) obtained from rice husk for the degradation of the polluting
dye Rhodamine B (RhB) through photocatalytic ozonation. The main objective
was to create an eco-friendly nanocatalyst from agro-industrial waste
(such as rice husk and lemon) and strategic Brazilian minerals (Nb
and Li), fostering a circular economy. Rice husk was used for SiO2 extraction, while LiNbO3 nanoparticles were biosynthesized
by using the hydrothermal method, employing lemon peel extract (Citrus latifolia). Characterization confirmed the
porous morphology and the creation of nanoparticles (34 nm) with a
high surface area, ideal for dye diffusion. Using central composite
rotatable design (CCRD), the SiO2/LiNbO3 system
showed high photodegradation efficiency (>97%) in less than 60
min,
adhering to the Langmuir–Hinshelwood kinetic model. Superoxide
radicals (O2
•–) and valence holes
(h+) were the main agents responsible for the degradation process.
The ecotoxicity of the final material was low. Furthermore, the use
of machine learning (ML) to anticipate the formation of degradation
intermediates highlighted the predictive ability of the K-nearest-neighbor
(KNN) model.