Experiences Using Mass Spectral Deconvolution For Groundwater Contaminant Screening

John A. Nichols (Portland) And Rodger B. Voelker (currently U. of Oregon, Eugene)
Oregon Department of Agriculture, Food Innovation Center

Abstract

GC/MS and AMDIS software is applied to detect unknown compounds in groundwater extracts. AMDIS (Automated Mass-Spectral Deconvolution and Identification System) was initially developed for verification of the Chemical Weapons Convention, and has been made available for free electronic downloading. The library of spectra that we searched against includes pesticides and other potential contaminants of interest to our client; and it is easily customized.

An SPE clean-up method with Oasis MAX cartridges achieves a 2000-fold concentration, for both neutral and acidic (phenoxy) compounds; the latter fraction is methylated with diazomethane.

A PTV large-volume injector allows 10 microliter injections, and full scan mass spectral data is collected. The extracts are clean enough that inlet maintenance is not unduly frequent. However, the long residence time in the PTV inlet does cause reduced sensitivity for some labile compounds. In general, method detection limits for typical pesticides range from 4 to 35 parts per trillion in groundwater. A few compounds show higher MDL's; certain volatile compounds showed low and erratic recoveries, suggesting volatility losses during the evaporation step prior to injection.

The AMDIS processing is very computationally intensive, as it looks at all potential "peaks", with and without various combinations of potentially interfering ions (e.g. column bleed). Potential matches receive a net score from 50 to 100, and other output is available. We print out all hits for evaluation. In our experience, scores higher than about 70 to 75 can almost always be confirmed! Match scores below 70 generally are not valid identifications. Diuron was a special case-any hint of it by GC meant that we followed up on it using LC/MS (which confirmed its presence in every instance).

In summary, we used AMDIS as a qualitative screening tool to supplement traditional target analysis for 27 pesticides. Nineteen contaminant compounds were observed among 100 well samples from Oregon's upper Willamette Valley, including 8 that we would have missed had we not used AMDIS. We gained considerable respect for its reliability. Data is presented showing "threshold of detection" levels to be very close to MDL's by our usual quantitative methods.