Today I was going back through my files and noted an “oldie but goodie” – a paper on background fitting for LIBS and Raman spectra by Igor Gornushkin and colleagues from 2003. Specifically: I. B. Gornushkin, et al. “Automatic Correction of Continuum Background in Laser-Induced Breakdown and Raman Spectrometry,” Appl. Spectrosc. 57, 197-207 (2003). They performed a flexible, piecewise analysis of each spectra, with an iterative algorithm that allowed them to properly fit the background automatically. Visually, the results were quite impressive.
They investigated the effects of the background algorithm on LIBS and Raman spectra. They found that for sparse LIBS spectra and for Raman spectra, the background removal was quite accurate – and in the Raman spectra it significantly improved the percentage of correct identifications of plastics, over non-corrected spectra. For line-rich LIBS spectra, the algorithm had a tough time, consistently over-predicting the background. In this case, there was no observed analytical improvement.
This paper reminds us that 1) algorithms (even pre-treatment algorithms) are important, and 2) that the proper algorithm for one situation or spectroscopy may not be the right answer in a different situation. These are fun problems to solve, but they take the proper tools and often take some time. A design of experiments approach with data analysis can often really help. If you’re interested in tools (like Analyze IQ), a consultation, or some advice on working with your spectral challenges, feel free to contact us!