Application of genetic algorithms in automated assignments of high-resolution spectra

Abstract

Abstract This paper describes an automated assignment and fitting procedure for high-resolution rotationally resolved spectra. The method is based on the application of genetic algorithms (GA) and both frequency and intensity information of these spectra is used. The basic ideas behind the GA technique is introduced and the particular fitness function critical for the success of the GA evaluation is discussed. The meta-optimization of the internal GA parameters for an optimal exploration of the error landscape of the spectrum fitting is investigated. A number of typical examples are given in which the use of automated spectrum assignments with the GA method is of crucial importance in the analysis. Examples are given for fits of very dense spectra due to overlap of a number of rovibronic spectra in conformers and isotopomers and strongly congested spectra in dimer systems. It is also shown that since the GA performs an overall fit of the complete spectrum very good information on the orientation of the transition dipole moments is gathered. Application of genetic algorithms in automated assignments of high-resolution spectraAll authorsW. Leo Meerts & Michael Schmitthttps://doi.org/10.1080/01442350600785490Published online:22 February 2007Table Download CSVDisplay Table Contents PAGE 1. Introduction 354 2. The genetic algorithm 357 2.1. Introduction357 2.2. The fitness function358 2.2.1. Definition358 2.2.2. Numerical evaluation of the fitness function359 3. Optimization of the GA fits 360 3.1. Effect of the width (Δ w) of the weight function w(r) on the convergence of the GA360 3.2. Meta-optimization of the internal GA parameters361 3.3. Uncertainties in and correlations between the parameters from a GA fit362 4. Methods 363 4.1. Experimental363 4.2. The calculated spectra364 4.2.1. The asymmetric rotor Hamiltonian364 4.2.2. Internal rotation Hamiltonian365 4.2.3. Intensities366 4.3. The structure determination367 4.4. The software used368 5. Examples of the success of GA automatic assignments 368 5.1. GA fits of very dense rovibronic spectra368 5.1.1. [7-D]phenol-N2 368 5.1.2. Benzonitrile-Ar369 5.1.3. 4-methylphenol and its water complex371 5.2. Simultaneous GA fits of a number of overlapping rovibronic spectra373 5.2.1. 7-azaindole (Pyrrolo[2,3-b]pyridine)374 5.2.2. Resorcinol (1,3-dihydroxybenzene)375 5.2.3. Conformers and isotopomers of tryptamine377 5.3. Strongly congested spectra in dimer systems379 5.3.1. Benzoic acid dimer revisited380 5.3.2. Benzonitrile dimer385 5.3.3. Phenol dimer387 5.4. Orientation of the transition dipole moment389 5.5. GA fit of rotational contours391 5.5.1. The FTIR spectrum of benzotriazole391 5.5.2. The low-resolution LIF spectrum of azaindole-water393 6. Summary 394 Acknowledgements 395 Appendix A 395 Appendix B 396 References 400 Supporting online material 403 Acknowledgments We would like to thank Jos Hageman, Ron Wehrens, Lutgarde Buydens and Gerrit Groenenboom for many helpful discussions. We thank Christian Ratzer, Grzegorz Myszkiewicz, Marcel Böhm, Ivo Kalkman, Chau Vu, Robert Brause, Violetta Bednarska and Marloes van Beek for their contributions to this work and Bert Groenenboom for giving the mathematical proof presented in Appendix A. The financial support of the Deutsche Forschungsgemeinschaft (SCHM 1043/9-4 and SFB663 project A2) is gratefully acknowledged. MS wishes to thank the Nordrheinwestfälische Akademie der Wissenschaften for a grant, which made this work possible. WLM thanks Timothy S. Zwier at Purdue University for his hospitality as this paper was being written. ` would like to thank the National Computer Facilities of the Netherlands Organization of Scientific Research (NWO) for a grant on the Dutch supercomputing facility SARA. This work was also supported by the Netherlands Organization for Scientific Research and the Deutsche Forschungsgemeinschaft in the framework of the NWO-DFG bilateral programme.