SCIENTIFIC PROGRAMME OF THE CONFERENCE

LIST OF ABSTRACTS

Keynotes and conclusions of the ROUND TABLE

EMSLIBS 2009 Round Table  - Tivoli Terme 30.09.09 17:30 – 19:05

Proposed title: How much LIBS applications are going to profit from the use of advanced chemometric tools?

Moderator: Roberta Fantoni

Introducing speakers:
Javier Laserna, Vincenzo Palleschi, Nikita Zorov, Ulrich Panne, Reinhard Noll, N. Konjevic

The discussion was very exciting, although large deviations from the proposed topic where encountered. Without entering in the details several points were arisen:

• The use of chemometrics tools may be necessary and unavoidable in the case of complex problems which cannot be solved only by LIBS spectra assignment. Typical examples are the determination of organic compounds by means of stoichiometry, deduced by LIBS, whenever it can be eventually hypothized. This is the case for instance of explosives standoff detection, whose identification may require also trace detection of minor elements.
• A large series problems related to spectral assignment, e.g. LIBS qualitative analysis, which are nowadays object of chemometrics, can be equally solved by smart use of simpler tools, such as combination of a few selected spectral ratio. In contrary chemometrics may be seriously needed for LIBS quantitative analysis, especially of trace elements in complex matrices.
• Although they are powerful method for statistic data analysis, none of the chemometric approach can solve those LIBS problems which are related to the sample nature (matrix effect, internal calibration) and related to the ablation process itself. To this respect, especially in LIBS, that differently from ICP collects the signal only from a small fraction of the ablated material, modeling of the ablation process and of the successive plasma evolution is still needed.
• Plasma parameters are necessary to the modeling and to establish the validity of LTE assumption utilized in quantitative data analysis. The need for a complete elemental data base, suitable to evaluate atomic and ionic parameters including the broadening coefficients, is heavily felt by the entire LIBS community. Such a data base should also include reliable partition functions for heavy and light elements, calculated in a large range of temperatures encountered in laser generated plasma for metallic and dielectric targets either in air or under vacuum (including the effects of lowering the ionization potential).
• Prior to the application of different chemometric tools for statistics, either sorting or quantitative analysis, a correct data processing procedure should be established, with proper normalization on internal standards or on other experimental results (multivariate recalibration, hyphenated techniques).
• Data fusion from different and complementary techniques could help the typical problems addressed by chemometric approaches. Methods for practical implementation still need to be defined in a statistically robust manner. In this respect the some ancillary data could be used for internal normalization.
• The choice of most appropriate chemometric tool should rely on the possibility to relate the data treatment procedure to the physical meaning of the variable selected with a major weight, To this respect PCA and PLS are preferable with respect to ANN analysis in which the physical meaning is completely lost.

In conclusion we may summarize the round table results in the following recommendations:

• Go back to fundamentals for sampling process modeling;
• Use reliable atomic data for quantitative data retrieval;
• Use appropriate internal standardization and/or calibrations prior to any data analysis;
• Do not use chemometrics as a black box, keep an eye on the physical meaning of the variable during the statistical analysis.