Chromatographic peaks are generally asymmetric owing to extra- and intra-column effects. This implies that the experimental retention time, t(R,p), and variance, σ 2/(p), do not properly describe the shape of a peak under the experimental conditions used, nor do they allow one to predict how this shape will change when the composition of the eluent is varied. Consequently, some of the conclusions drawn from these experimental parameters may contain an extent of error that cannot be evaluated. This may occur for instance when some of the lower order statistical moments are used to estimate thermodynamic properties related to the solute-eluent- chromatographic column system. The aim of this paper is to test whether and to what extent information can be improved in ion chromatography when the moments estimated by some suitable fitting functions are compared to the experimental ones. The following fitting functions were tested: the exponentially modified Gaussian (EMG) (the peak deconvolution leads to the Gaussian curve which would be obtained in the absence of distortion), the bi- Gaussian (bi-G) (the deconvolution leads to the sum of two half-Gaussians), and the two-Gaussians (2-G) (the deconvolution leads to two Gaussians). The last one is proposed in this paper for the first time. Once we had evaluated how each component of a modern IC chromatograph (namely loop, column, post- column conditioning module and detector) affects the peak shape, the three fitting models were tested: (i) to compute β 1, and β 2 stability constants for the Cd +2/Cl - system from the measurement of the first-order moments at a variable Cl - concentration and constant ionic strength (0.22). It was found that: β 1 =28.86 (EMG), 28.53 and 28.59 (2-G), 28.99 (bi-G) and β 2 = 57.96 (EMG), 59.01 and 58.98 (2-G), 53.92 (bi-G). The β values calculated from the experimental retention times referred to the peak maximum were: β 1 =27.89 and β 2 =64.95; (ii) to evaluate, for the various solutes, the dependence on the first order moment (m 1) of some parameters of the i fitting functions, such as τ (the time constant of the exponential modifier in the EMG), σ(i) (the square root of peak variance) and the asymmetry a/b (graphically measured as the ratio of the distances of the peak contour from the abscissa of peak maximum, at 0.1 peak height). A linear dependence on m 1 was found for: σ(i) with a slope which depends on the fitting model used and not on the type of solute: τ and a/b with a slope which depends on the solute type. These findings are of interest: (i) to provide a sound tool for the accurate IC estimate of stability constants based on the Gaussian first order moment computed from the deconvolution of the experimentally distorted peak (ii) to evaluate how the shape of the solute peak changes with m 1 when the shift of the peak along the time (or volume) scale due to variations in the eluent composition can be thermodynamically predicted. In the last case the optimized peak separation conditions, which take the effect of peak shape variation into account, might be more efficiently found.
Enhancing the quality of information obtained by a comparison between experimental and deconvolved peak parameters in ion chromatography
CECCARINI, ALESSIO;
1997-01-01
Abstract
Chromatographic peaks are generally asymmetric owing to extra- and intra-column effects. This implies that the experimental retention time, t(R,p), and variance, σ 2/(p), do not properly describe the shape of a peak under the experimental conditions used, nor do they allow one to predict how this shape will change when the composition of the eluent is varied. Consequently, some of the conclusions drawn from these experimental parameters may contain an extent of error that cannot be evaluated. This may occur for instance when some of the lower order statistical moments are used to estimate thermodynamic properties related to the solute-eluent- chromatographic column system. The aim of this paper is to test whether and to what extent information can be improved in ion chromatography when the moments estimated by some suitable fitting functions are compared to the experimental ones. The following fitting functions were tested: the exponentially modified Gaussian (EMG) (the peak deconvolution leads to the Gaussian curve which would be obtained in the absence of distortion), the bi- Gaussian (bi-G) (the deconvolution leads to the sum of two half-Gaussians), and the two-Gaussians (2-G) (the deconvolution leads to two Gaussians). The last one is proposed in this paper for the first time. Once we had evaluated how each component of a modern IC chromatograph (namely loop, column, post- column conditioning module and detector) affects the peak shape, the three fitting models were tested: (i) to compute β 1, and β 2 stability constants for the Cd +2/Cl - system from the measurement of the first-order moments at a variable Cl - concentration and constant ionic strength (0.22). It was found that: β 1 =28.86 (EMG), 28.53 and 28.59 (2-G), 28.99 (bi-G) and β 2 = 57.96 (EMG), 59.01 and 58.98 (2-G), 53.92 (bi-G). The β values calculated from the experimental retention times referred to the peak maximum were: β 1 =27.89 and β 2 =64.95; (ii) to evaluate, for the various solutes, the dependence on the first order moment (m 1) of some parameters of the i fitting functions, such as τ (the time constant of the exponential modifier in the EMG), σ(i) (the square root of peak variance) and the asymmetry a/b (graphically measured as the ratio of the distances of the peak contour from the abscissa of peak maximum, at 0.1 peak height). A linear dependence on m 1 was found for: σ(i) with a slope which depends on the fitting model used and not on the type of solute: τ and a/b with a slope which depends on the solute type. These findings are of interest: (i) to provide a sound tool for the accurate IC estimate of stability constants based on the Gaussian first order moment computed from the deconvolution of the experimentally distorted peak (ii) to evaluate how the shape of the solute peak changes with m 1 when the shift of the peak along the time (or volume) scale due to variations in the eluent composition can be thermodynamically predicted. In the last case the optimized peak separation conditions, which take the effect of peak shape variation into account, might be more efficiently found.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
