Presented at the 1983 Pittsburgh Conference
DATA PROCESSING AS AN ALTERNATIVE TO EXTENSIVE SAMPLE CLEANUP IN THE GAS CHROMATOGRAPHIC DETERMINATION OF PCBs
David Gossman SYSTECH Corporation
David Gossman is now with Gossman Consulting, Inc.
The analysis of oils and solvents for PCBs is long and costly primarily because of the extensive sample cleanup work generally needed. The use of a computer to "clean up" the data instead of chemicals to clean up the sample can present substantial time and dollar savings in this determination and others like it, such as pesticides.
The greatly simplified sample preparation consists of dilution in an appropriate solvent, drying over anhydrous sodium sulfate, and centrifuging or filtering to remove any particulates. This takes 2 to 5 minutes.
A Perkin-Elmer Sigma 115 is used for a temperature-programmed run. The electron capture detector output is taken as slice data by the Sigma 15 computer. These data are then transferred to the Perkin-Elmer 3600 data station for actual data processing and analysis.
The data for the run is then graphically compared with a series of PCB standards stored on floppy disk. The analyst decides which standard is closest and also chooses what portion of the chromatographic time line will be used for further data processing.
The data are then fitted using a least squares technique, and the fitted data are plotted on the graphics screen. Also plotted is a difference spectra and a line representing the standard deviation. This information is then used to filter out peaks that exist in the sample but not in the standard. The lease squares routine is then repeated, using only the good data. Another fit is plotted and the whole process may be repeated, if necessary. Each time the data is plotted, numerical value for the factor, baseline, standard deviation, and correlation coefficient are also printed on the screen. Finally, at any point during this process, a hard copy of the screen can be obtained for documentation purposes.
This process has proven highly reliable on all but the dirtiest samples submitted for analysis. In addition, the statistical data provided using this method is very helpful in calculating the accuracy of a given determination.