Predictive Insights from Machine Learning-Assisted Laser-Induced Breakdown Spectroscopy of Lanthanum Substituted Bismuth Ferrite

The co-precipitation technique successfully synthesized lanthanum (La³⁺) substituted bismuth ferrites (BiFeO₃ or BFO) with chemical formula Bi_1−xLa_xFeO₃ (0.0 ≤ x ≤ 0.075). X-ray diffraction analysis unveiled a rhombohedral distorted perovskite structure for BiFeO₃ with space group R3c. Notably, an increase in La³⁺ concentration correlated with a rise in the average crystallite size, from 16 to 41 nm. The scanning electron microscopy images depicted a non-uniform spherical morphology. Fourier transform infrared spectroscopy confirmed the perovskite structure of BiFeO₃, with metal-oxide bonds evident in the wavenumber range of 492–538 cm⁻¹. UV–visible spectroscopy revealed a reduction in the energy band gap from 3.17 to 2.77 eV as the concentration of La³⁺ increased. The LIBS analysis identified the presence of bismuth (Bi), iron (Fe), and lanthanum (La) in the samples. To validate the local thermodynamic equilibrium, the McWhirter criteria were utilized. Employing principal component analysis alongside LIBS spectra proved effective in classifying materials with minimal concentration variations. The proposed ML models for LIBS spectroscopic data are principal components analysis, discriminant analysis (DA), support vector machines, and neural networks. DA showed better performance as compared to other models. Our results align with the experimental findings, affirming the credibility of the model.