Investigations on the resolution of overlapped chromatograms

Abstract

High-Performance Liquid Chromatography (HPLC) and Gas Chromatography (GC) are generally used for complex chemical analysis. Some components in the chemicals produce overlapping chromatograms due to their similar retention time (RT) and hence difficult to analyze. Many of the curve fitting deconvolution methods used so far required many peak parameters as an input to get accurate deconvolution. They were time consuming and required human intervention. The multivariate curve resolution techniques developed so far were not accurate for the separation of severely overlapped chromatograms. Even if the peaks were moderately overlapped, some of the already developed algorithms failed to resolve. The Non-negative Matrix Factorization (NMF) based separation algorithms did not give unique results. The convergence issues of the NMF algorithms were also questionable. Hence studies were initiated to separate the chromatograms of acetone and acrolein mixture by proposing two curve fitting algorithms viz., substitutive algorithm and iterative curve fitting algorithm. The efficiency of the algorithms has been tested under real time and three different simulated cases viz., partially overlapped, severely overlapped and embedded chromatograms. The proposed algorithms prove to be user friendly with less human intervention and less separation time. Both substitutive and iterative algorithms are effective for simulated partially overlapped chromatograms only. Further, the proposed iterative curve fitting algorithm is more efficient in handling chromatograms with tailing peaks. However the proposed algorithms are not compared with the existing curve fitting methods as the focus of the study is to resolve strongly overlapped chromatograms in a simpler way. The severity of overlap exerts limitation in the use of the above types of curve fitting algorithms. Hence, the potentiality of existing multivariate curve resolution technique namely MCR-ALS (Multivariate Curve Resolution-Alternating Least Square) method have been tested