LAboratoire de Spectrochimie Infrarouge et Raman – UMR 8516
  • Français
  • English
  • Spanish


Chemometrics is concerned with the application of mathematical and statistical data analysis methods for the characterization, modeling and understanding of complex chemical samples, systems and processes.

Multivariate resolution of unknown spectroscopic mixtures is our main interest at fundamental and applied levels. The aim is to unravel pure species components providing their kinetic/spatial evolutions and spectral features.

From an analytical perspective, multivariate calibration models are being built for the prediction of quantitative or qualitative information from complex (industrial, naturals, …) sample spectra. Among the different research lines investigated are data preprocessing and variable selection.

On another hand, raw spectroscopic data may require dedicated data processing to handle instrument-specific data structures, as in ultrafast time-resolved spectroscopy, or hyperspectral imaging.

Methods are also developed for simultaneous analysis or fusion of spectroscopic data from multiple experiments providing complementary information. This is of particular relevance for the investigation of complex evolving systems and reactional schemes.

Chemometrics for the resolution of process spectroscopy data (reactions, images, evolving chemical systems)


• Time-resolved spectroscopy (rapid-scan FTIR, UV-Vis and IR femtosecond transient absorption)
• Hyperspectral imaging (Raman, IR, NIR, RPE)

Photoreactivity: spectrokinetic characterization of transient species in femtosecond transient absorption spectroscopy

Investigation of the photophysics of benzophenone. Multivariate curve resolution of femtosecond transient absorption spectroscopy data. Transient spectra and kinetics of the extracted species.
(Ref. P 2008-06)

Study of hydration shell of solvated molecules. Direct characterization of water / ethanol mixtures with terahertz spectroscopy into a microfluidic system. Development of chemometrics tools in order to obtain with no a priori a simultaneous extraction of concentration profiles and corresponding spectra of all pure compounds present in the chemical system. Major achievement: detection of two layers defining the hydration shell.
Aerosols, cell biology: molecular identification and chemical map extraction of all pure coumpounds

Infrared spectroscopic imaging with a synchrotron beamline, single cell characterization (human cancer cell, HeLa type). Multivariate curve resolution method is a good way to extract chemical maps and pure spectra in such complex samples

Multivariate calibration for qualitative and quantitative analysis

Industrial collaborations:

Total, Michelin, IFEN, Roche


NIR, IRTF, Raman

Predicting the composition of textile products using near infrared spectroscopy. (ref. P 2008-21)

Chemometrics methods can be used to discriminate between complex samples and relate their class-membership to characteristic spectral features. The results shown illustrate the discrimination between mutant and wild cells of vegetal roots.

Data preprocessing and variable selection

Genetic algorithms are methods to find optimal solutions when too many influential parameters have to be explored. In this case, classical exhaustive search methods are effectively overwhelmed by combinatory explosion. The study presented here shows how it is possible to obtain a simultaneous optimization of wavelength selection and spectral preprocessing for the development of a chemometrics model capable of predicting concentrations of interest from near infrared spectra. (Ref. P 2011-16)

Statistical methods, signal processing

Data processing to correct for spectrokinetic distorsion attributed to stimulated Raman amplification artifacts in femtosecond transient absorption spectroscopy. (Ref. P 2011-42)

Methods and data structures

The superresolution concept

Main objective: push the limits of spatial resolution in spectroscopic imaging. The simultaneous use of several low resolution images of a same object (observed from different point of view) allows us to generate one higher resolution image. Presented case: submicronic Raman imaging of dust particules.

Dealing with non-gaussian noise in time-resolved spectroscopy

PARAFAC decomposition of multiway data. The data structure is assembled gathering noise matrices obtained repeating time-resolved spectroscopy experiments. (Ref. P 2009-12)