Error analysis in non-linear modelling of HPLC retention

Short description:

Basically, there is two ways to elute compounds in HPLC, in isocratic (the mobile phase composition does not change) and in gradient (by increasing gradually the elution strength of the mobile phase). This project is about a particular aspect of modelling HPLC gradients. It happens that it is difficult to establish the error propagation in HPLC modelling when there is transference between elution modes. In other words, it is useful to check how does the error propagation behaves in situations in which data is collected in isocratic but used to predict gradients (transference from isocratic to gradient), or viceversa (data collected in gradient elution mode to predict isocratic). The difficulty is that gradient models (when the parabolic equaion is used or when the gradient is not linear) do not have an analytical solution.

We inspected the application of the error propagation of both cases (from isocratic to gradient and vice-versa) by studying the Jacobian matrices associated to the regression. We could find that, when gradient retention is predicted out of isocratic experiments, we can define a concentration range of the mobile phase in which obtaining the correct prediction is of paramount importance. Consequencly, we should use weighted regression to compensate for this effect. Details about this work can be found elsewhere [7a10a].


This project was developed at the University of Valencia (Spain). Several people were involved. See authors of the publications for more details about authorship.


University of Valencia


None available


None available.