Application of artificial neural networks and genetic algorithms to modeling molecular electronic spectra in solution

A novel approach is presented for finding the vibrational frequencies, Fran ck-Condon factors, and vibronic linewidths that best reproduce typical, poo rly resolved electronic absorption (or fluorescence) spectra of molecules i n condensed phases. While calculation of the theoretical spectrum from the molecular parameters is straightforward within the harmonic oscillator appr oximation for the vibrations, "inversion" of an experimental spectrum to de duce these parameters is not. Standard nonlinear least-squares fitting meth ods such as Levenberg-Marquardt are highly susceptible to becoming trapped in local minima in the error function unless very good initial guesses for the molecular parameters are made. Here we employ a genetic algorithm to fo rce a broad search through parameter space and couple it with the Levenberg -Marquardt method to speed convergence to each local minimum. In addition, a neural network trained on a large set of synthetic spectra is used to pro vide an initial guess for the fitting parameters and to narrow the range se arched by the genetic algorithm. The combined algorithm provides excellent fits to a variety of single-mode absorption spectra with experimentally neg ligible errors in the parameters. It converges more rapidly than the geneti c algorithm alone and more reliably than the Levenberg-Marquardt method alo ne, and is robust in the presence of spectral noise. Extensions to multimod e systems, and/or to include other spectroscopic data such as resonance Ram an intensities, are straightforward. (C) 2001 American Institute of Physics .