This work describes the use of a quantified maximum entropy method for the optimisation of analytical information from EPR spectra. EPR spectra are normally recorded as first derivatives of absorptions and, like all experimental data, are distorted from the ideal lineshape by a combination of systematic and instrumental factors. The first and crucial step in the analysis of a spectrum by maximum entropy methods is the definition of a function --- the point spread function (PSF) which represents the spread of the data around a point, or the distortion from ideality. For EPR spectra, the PSF is selected by differentiating a lineshape comprising Gaussian, Lorentzian and square wave contributions. Using this inital estimate of the PSF, the most probable result is calculated as the frequency disitribution of intensities which, convolved with the PSF, gives the best fit to the experimental data. Discrepancies between calculated and experimental data are used to guide the calculation by an iterative procedure and exploration around the optimum result allows the determination of the error associated with each line's frequency and intensity. The ability to use composite PSFs lends further refinement, either for the disentangling of multiline spectra or for the determination of the intensity of weak spectra of known radicals, a problem often encountered in biological EPR spectroscopy.
Quantified maximum entropy reconstruction of the complex multiline EPR spectrum of paraquat allows the accurate determination of the 4 hyperfine splittings. The approach is then used on a system involving unknown radical species formed by crushing lettuce in the presence of a spin trap. This reconstruction reveals the presence of at least 2 radicals and the parameters derived suggest that one of them is a hydroxyl radical adduct --- a result not obtainable by inspection.