The problem has to do with the underlying FT-PGSE data analysis, which is
probably based on Provencher's CONTIN routine or similar.
It is well known that you cannot really deconvolute overlapping
exponentials particularly well. CONTIN does a good job, but there is not
enough information in the data.
Probably Bruker programmers have set one of the fitting parameters in
CONTIN to a 'smooth' value. Then what you get is smooth and broad
distributions. With a 'sharper' value, you get all kinds of strange
distributions of (in this application) diffusion coefficients.
Neither of these have any physical significance, since there is just not
enough information ! The CONTIN approach to FT-PGSE data processing treats
each frequency individually, and each such fit has the same problem.
The final 'DOSY display' (DOSY is really just a display mode for FT-PGSE,
nothing else) is here just based on interpolated results from individual
frequency channels. Of course, the final result will suffer from the same
problem.
A better approach in this context is to really make use of all of the
available information in the data set. This can be made through multivariate
methods, like NIPALS (by Kubista) or GRAM or DECRA (developments of the
Kubista approach by Antalek et al), or through my own CORE approach (a global
least-squares one). Here, the data for individual frequencies are no longer
treated independently. If desired, the final results can be summarized in a
DOSY display mode.
See e.g. JMR 135 (1988) pp 236- for some relevant references to these methods.
Peter Stilbs
*********************************************************************
Peter Stilbs, Prof. tel: +46 8 7908201
Physical Chemistry fax: +46 8 7908207
Royal Institute of home phone: +46 8 54023367
Technology (KTH) mobile phone: +46 70 6220736
SE-100 44 Stockholm mailto:peter@physchem.kth.se
Sweden http://omega.physchem.kth.se/~peter/
"Politics is perhaps the only profession for which no preparation is
thought necessary" (Robert Louis Stevenson (1850-1894))
*********************************************************************