Enhancing the Limit of Detection for Comprehensive Two-Dimensional Gas Chromatography (GC×GC) using Bilinear Chemometric Analysis

Abstract

The chemometric method referred to as the generalized rank annihilation method (GRAM) is used to improve the precision, accuracy, and resolution of comprehensive two-dimensional gas chromatography (GC x GC) data. Because GC × GC signals follow a bilinear structure, GC x GC signals can be readily extracted from noise by chemometric techniques such as GRAM. This resulting improvement in signal-to-noise ratio (S/N) and detectability is referred to as bilinear signal enhancement. Here, GRAM uses bilinear signal enhancement on both resolved and unresolved GC x GC peaks that initially have a low S/N in the original GC × GC data. In this work, the chemometric method of GRAM is compared to two traditional peak integration methods for quantifying GC x GC analyte signals. One integration method uses a threshold to determine the signal of a peak of interest. With this integration method only those data points above the limit of detection and within a selected area are integrated to produce the total analyte signal for calibration and quantification. The other integration method evaluated did not employ a threshold, and simply summed all the data points in a selected region to obtain a total analyte signal. Substantial improvements in quantification precision, accuracy, and limit of detection are obtained by using GRAM, as compared to when either peak integration method is applied. In addition, the GRAM results are found to be more accurate than results obtained by peak integration, because GRAM more effectively corrects for the slight baseline offset remaining after the background subtraction of data. In the case of a 2.7-ppm propylbenzene synthetic sample the quantification result with GRAM is 2.6 times more precise and 4.2 times more accurate than the integration method without a threshold, and 18 times more accurate than the integration method with a threshold. The limit of detection for propylbenzene was 0.6 ppm (parts per million by mass) using GRAM, without implementing any sample preconcentration prior to injection. GRAM is also demonstrated as a means to resolve overlapped signals, while enhancing the S/N. Four alkyl benzene signals of low S/ N which were not resolved by GC x GC are mathematically resolved and quantified.