Determination of enantiomeric composition of ibuprofen in solid state mixtures of the two by DRIFT spectroscopy

The enantiomeric purity of ibuprofen was determined in a simple manner by diffuse reflectance infrared Fourier transform (DRIFT) spectroscopy with artificial neural networks (ANNs) methodology. A series of 17 binary mixtures was created using different proportions of the two enantiomeric forms. Sample mixtures were dispersed as a 5% (w/w) mix in KBr, and spectra were measured immediately after mixing. The original spectra were sampled between 650.16 and 3999.40 wavenumbers (cm(-1)) and reduced to 1738 spectral intensities during the data collection. These spectral data were further processed to smooth the noise in the spectrogram. Reduction and transformation of the input data enhanced the ANN performance. The 1738 reflectances were reduced to 173 averaged spectral values, each from 10 consecutive wavenumbers. Two ANNs models with one or two hidden layers were trained, tested and validated. Both models had 173 averaged spectral values as the inputs and two output neurons, one for the percentage of each ibuprofen enantiomer. The number of hidden layers and hidden neurons was optimized. The ANN model was built by comparing the predictions obtained from several high scoring models. The best results were obtained with two hidden layers having six hidden neurons in each layer. The method is highly sensitive and precise. A working range of 1-100% of the R(-)-enantiomer present as an impurity in S(+)-enantiomer was established with a minimum quantifiable level (MQL) of 1.67% and a limit of detection (LD) of 0.5% (w/w). The average+/-S.D. recovery values were 100.95+/-1.82 and 98.02+/-4.84 for R(-)- and S(+)-enantiomer, respectively. (C) 2000 Elsevier Science B.V. All rights reserved.