US EPA

2184880 

Journal Article 

Least-squares analysis of gas-chromatographic data for polychlorinated biphenyl mixtures 

Spencer, JL; Hendricks, JP; Kerr, D 

1993 

Yes 

Journal of Chromatography A
ISSN: 0021-9673
EISSN: 1873-3778 

654 

1 

143-149 

English 

WOS:A1993MW22300015 

A least squares method for the analysis of gas chromatographic data for polychlorinated biphenyl (PCB) mixtures was developed and tested. The data analysis involved two stages. First, the peak area versus retention time was used to reconstruct a continuous function of time that approximated the original signal obtained from the chromatograph detector as the sample was eluted from the column. This function was then sampled at evenly spaced times to produce a vector. Assuming the chromatograph operated at a linear range, the signal produced was a linear combination of the signals produced by pure samples of each of the PCBs in the mixture. The coefficients of the linear combination were then proportional to the amounts of pure PCB which made up the sample being analyzed. Second, the coefficients were determined by requiring that they produced a least squares fit to the chromatogram. Formulae pertaining to elution curve reconstruction and parameter estimation were presented. Retention times for Arochlor-1242, Arochlor-1254, and Arochlor-1260 formed the standards. While Arochlor-1242 had large peaks at lower retention times, the other two yielded peaks over a wide range of retention times. Original and reconstructed chromatograms for Arochlor-1260, and reconstructed chromatograms for all three compounds were presented. Thirteen blends of the three PCBs were tested. In general, agreement was excellent. The maximum difference between known and calculated mass percents was 1.89%, and the average difference was 0.26%. In all cases, the standard error lay between 0.10% and 0.40%. The authors conclude that the least squares procedure yields results quite close to the known compositions of PCB blends and that the optimization procedure described allows the determination of the best retention time factor.