What is it about?
Analyzing the composition of fragrance mixtures currently relies on chromatography (slow and expensive) or human smell panels (subjective). We used femtosecond thermal lens spectroscopy (FTLS) to measure optical signatures of individual fragrance ingredients, then built statistical models to predict mixture compositions from their combined spectra. FTLS works by focusing a short laser pulse into a sample and measuring how the resulting heat changes the refractive index. Each fragrance compound produces a characteristic thermal lens signal. By measuring pure ingredients first, we can decompose a mixture's signal into its constituent contributions. We tested this on binary and ternary fragrance accords, achieving quantitative composition estimates without separating the components. The measurement takes seconds rather than the hours required by chromatographic methods.
Featured Image
Photo by MeSSrro on Unsplash
Why is it important?
Quality control in the fragrance industry requires verifying that blends match their target compositions. Current methods are either slow (gas chromatography), expensive, or subjective (sensory panels). A rapid optical measurement that gives quantitative composition estimates would fit naturally into production workflows. FTLS has the additional advantage of being non-destructive and requiring only small sample volumes. The statistical framework we developed (inverse modeling from thermal lens signals) generalizes to other multi-component systems beyond fragrances. This paper reports the results of a seven-year measurement campaign across multiple laser setups, providing a substantial dataset for validating the approach.
Perspectives
This work spans seven years of intermittent measurements across different laser configurations. My initial approach used traditional curve-fitting, but the thermal lens signals from mixtures did not decompose cleanly that way. Switching to a statistical inverse-modeling framework solved the problem. Instead of fitting physical parameters, we trained models on pure-component measurements and used them to predict compositions. The approach is simple but effective, and it avoids the assumptions about mixture behavior that made curve fitting unreliable. The long timeline reflects the experimental reality of collecting reproducible femtosecond measurements across different setups. Each dataset had to be cross-validated before we could pool them.
Rohit Goswami
University of Iceland
Read the Original
This page is a summary of: Compositional Analysis of Fragrance Accords Using Femtosecond Thermal Lens Spectroscopy, Chemistry - An Asian Journal, June 2025, Wiley,
DOI: 10.1002/asia.202500521.
You can read the full text:
Contributors
The following have contributed to this page
