Inconsistency of chili() and exchange() functions

[katarkon]

I try to simulate the thumbling effect by three-site chemical exchange using exchange() function.
It would seem that chili() and exchange() functions use the same method, so they must give similar results, at least in fastmotion regime. However, there are some inconsistencies between the results. Firstly, slight shift of the spectra is observed. Secondly, relatively similar line shapes are observed when rate constant and correlation time are linked by strange expression k=0.784e-6*tcorr, whereas the coefficient 1e-6 should be expected.
Is there any explanation of such inconsistency?

[Stefan Stoll]

chili() and exchange() use fundamentally different approaches. chili() uses an eigenfunction expansion technique with a basis set that can be changed in Opt.LLKM. When you compare this to a multi-site jump model, the expected results are going to different, esp. when you compare chili with a exchange model with only 3 sites. In a multi-site picture, chili is equivalent to an jump model with infinitely many sites (assuming the basis set is large enough that spectrum is converged).