Answers are grounded in this paper and cite their sources.
When overlapping peak areas in gas chromatograms are quantitatively resolved by either graphical methods or current methods of numerical analysis, gross inaccuracies frequently result. Sometimes, the problem can be circumvented experimentally with columns that provide better separations but such improvements of column resolution often entail excessive retention times. This prolongs the duration of an analysis, and it reduces the sampling rate at which continuous processes can be monitored. A numerical method is proposed here which can accurately resolve overlapping peak chromatograms without any change of apparatus. The resolutions derived from the proposed method are exact providing that the detector response at any time is proportional to the concentration of each component and that the response to two or more components is equal to the sum of individual responses. The proposed method is free of arbitrary assumptions concerning either the functional form or the derivatives of component response curves. The information that is required to resolve peak areas by the proposed method is derived from reference samples of nonoverlapping components whose chromatograms span the records of unknown samples. The resolution of closely spaced and overlapping peaks can then be done far more accurately than by graphical techniques or other forms of numerical analysis. Reasonably accurate results are also obtained with partially missing samples records due to peaks of concentrated components which go off scale and are truncated. Less basic problems pertaining to specific applications, such as automatic peak identification and grouping of peaks for fitting, are potentially soluble by the proposed method but they are not discussed in the paper. For the purposes of automating the proposed method and carrying it out faster and more accurately on a digital computer, the Gram-Schmidt orthogonalization procedure is developed in sufficient detail so that the underlying algorithm can be readily programmed. The problem of estimating base line drift is handled with the “minimum-energy” curve fitted by spline interpolation. An error control procedure based on the Chi-square distribution is suggested for detecting systematic errors in the presence of random experimental errors. The proposed method of numerical analysis and the auxiliary computer techniques appear applicable to a wide class of peak resolution problems not only in GC and LC but also in absorption, emission, scattering, and other spectra. A notable exception to the applicability of the proposed method is in thermal conductivity detection where the response at any time to two or more components is not a simple sum of individual responses, but depends instead upon the composition of the mixture.
No takes yet. Share an insight, caveat, or question.
B. Goldberg (1971) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: