Possibilities of retention modeling and computer assisted method development in supercritical fluid chromatography

On the inherent data fitting problems encountered in modeling retention behavior of analytes with dual retention mechanism

“…However, we already experimentally obtained that a simple model such as the Neue-model can successfully be used for gradient prediction based on gradient scouting runs [19]. Of course, the modeling should be extended if one wants to predict retention at different flow rates/column lengths (thereby changing the pressure).…”

“…However, the hands-on chromatographist who is developing methods, is only interested in the retention as a result of both. In that regard, this study (and the previously published studies [18,19]) investigate the possibilities of retention modeling using simple models based on readily available information (fractions of MeOH and corresponding retention times). The authors agree that once it is better understood how the retention is affected by the pressure, the retention models should be extended to include this pressure dependence.…”

“…To conduct these experiments both in HILIC and in SFC, and, for poorly and highly retained compounds is practically almost not feasible. Moreover, although the data generation is performed in a pure mathematical way, the resulting isocratic retention curves were very similar to those obtained in previous studies [18,19]. As a justification for this statement, we included some experimental isocratic HILIC and SFC data (Supplementary Material, Fig.…”

“…Indeed, retention in SFC is the result of both the mobile phase composition and the pressure. These pressure effects might explain why better prediction accuracy was obtained using gradient scouting runs vs. isocratic scouting runs [19]. Extending Eq.…”

“…frequently used models proposed in literature to cope with this non-linear retention behavior are the following [14][15][16][17][18][19]: ln(k) = ln(k w ) − S 1 ϕ − S 2 ln(ϕ) [Mixed model] ( 3) ln(k) = ln(k w ) + 2 · ln(1 + S 2 ϕ) − S 1 ϕ 1 + S 2 ϕ [Neue-model] (4) where ϕ is the fraction of water (in the case of HILIC) or MeOH (in the case of SFC). The present study originated from the general observation that it appears to be much more difficult to accurately predict retention times in HILIC [18] and SFC [19] than in RPLC and NPLC. These difficulties already start in the model fitting stage, where sometimes severe convergence problems are encountered.…”