| Abstrakt: |
An inhomogeneous lineshape model based on a convolution of Gaussian and Lorentzian functions is very common in EPR spectroscopy. For convenience, the convolution integral often is approximated by a sum of Gaussian and Lorentzian functions. A more accurate and general approach to linewidth simulation and fitting by use of a fast convolution algorithm is described. A Levenberg-Marquardt method is applied in the multidimensional optimization of simulation parameters. The partial-derivative matrix required for the Levenberg-Marquardt procedure is calculated without digital differentiation. To account for a microwave shift observed in 94 and 1.1 GHz EPR experiments, a dispersion parameter is included in the fitting function. All fitting parameters, including Lorentzian and Gaussian components to the linewidth, microwave phase shift, resonance position, and intensity, are extracted from experimental spectra directly. A fixed and/or variable superhyperfine pattern (or envelope function) may be included in the fitting procedure without a significant increase in computational time. Three examples of this technique are shown: extraction of Lorentzian broadening for a nitroxide EPR spectrum at various oxygen concentrations, penetration of a 1.1 GHz microwave field into a lossy dielectric sample, and processing of a mixed absorption-dispersion nitroxide 94 GHz EPR spectrum in the fast-motion limit. |
Full Text from ScienceDirect