Exponential corrections can be accomplished in different ways (quadratic calibration line, layer thickness corrections), but they usually do not apply in Cement based material analysis. What you probably want to do is correct for matrix effect: use the FP model (make sure that the sum of the compounds in your standards and application adds up to 100%) or the Classic model (but add the alpha's by going through the popup that is called by pressing the 'theory' button). For detailed help (or a general XRF course) you can contact the application specialists from PANalytical. Useful link: http://www.panalytical.com/Home.htm for world wide contacs and course program.

Thanks for your comments, can it be used for cement though?, I saw that the RMS was reduced by 30% for CaO, and my calibration curve performs better.

My own experience as a cement chemist shows that a calibration using the classic model with alphas computed for median concentration values of your product using 1:10 dilution gives excellent results. If your application is set up correctly then you should find sample preparation is the limiting factor with respect to reproducibility and precision. I did find that for cement standards I got improvement if I corrected the the reccomended values for the measured LOI of the CRM at time of bead preparation. My experience was that FP gave better results over a wide concentration range but best ultimate precision and accuracy for a given matrix was acheived with classical calibration with fixed alphas calculated. The above expressed opinions were validated by assessment of LTIH PTP results and NIST CRM's. Your calibrations should definitely be linear. There are instances where extreme variations in major compenents may require beta corrections to linearise the calibration curve (as for chalcopyrite/bornite ore complex for example) but this is the exception. Cement clinker systems are chemically well defined and stable and should present no problem. My personal opinion is that XRF obeys exactly the laws of physics and I would counsel against using any correction which you cannot justify by the laws of physics or the practical application of spectroscopy (interference by line overlap for example) i.e. you should calibrate with understanding. Remember that sample preparation is part of the calibration.