Noise Reduction in Laser-Induced Breakdown Spectroscopy (LIBS) Analysis ​​​

Laser-induced breakdown spectroscopy (LIBS) enables rapid, multi-element identification but degrades on low-SNR, misaligned, or aging instruments. Driven by the goal of expanding access to reliable, interpretable spectroscopy for schools and small labs, we present a fully composable, instrument-agnostic pipeline that couples dual continuum removal (arPLS + rolling percentile), adaptive template synthesis with FWHM sweep, phase-correlation micro-alignment, and non-negative-least-squares-based (NNLS) detection scored by fraction-of-variance-explained (FVE) with band-hit validation. Tested on more than 100 spectral sets spanning coins, alloys, and soils, the approach yields visibly cleaner fits and substantially better detection at low false-alarm rates than traditional peak-picking or fixed-template methods. Future work will validate the pipeline in progressively more complex matrices—starting with controlled multi-element mixtures at known concentrations and advancing to complex organic and aqueous media—culminating in seawater microplastics and mangrove heavy-metal assays to quantify limits of detection and interference. The implementation is open, modular, and runs end-to-end in < 0.6s per sample; we aim to lower the barrier to quantitative spectroscopy for under-resourced research settings.