7b. As an alternative to the method outlined above, we performed experiments with different number of T2-filters employed within the τ2 evolution period in the PGSTE experiment (see Fig. 2). As shown in Fig. 8, the experiments qualitatively performed as expected and yielded steeper and steeper diffusional decays with increasing number of filters applied. To obtain a clearer quantitative picture, in Fig. 9 we present the diffusion coefficients

obtained by fitting the classical Stejskal–Tanner expression to data obtained by different number of T2-filters embedded into the stimulated echo sequence as given by Fig. 2. Without any filter applied, the apparent diffusion coefficient Dinaciclib was much lower than the true Df value. Just a few filters

provided a rough correction for this effect and, as illustrated with the Δ = 10 ms data, the obtained diffusion coefficients seemed to converge to a constant level when τex became comparable to the inverse of the exchange rate, a behavior predicted in Fig. 4a. One should also note that diffusion coefficients obtained with different Δ but with the same τex are, within error, identical. It is important to note that the Df values obtained by the method proposed here seem to be close but significantly above the corresponding value obtained by performing the well accepted correction method [4], [8] and [37], that is estimating the exchange rate by the Goldman–Shen experiment and correcting for the effects of the exchange by inserting that exchange rate into the Kärger model. To our opinion, this difference highlights that the exchange-suppression method has the capacity to provide find more more accurate diffusion data. Namely, the correction method has several shortcomings. First, as has been thoroughly discussed and also demonstrated by simulations for unless specific cases [49], the Kärger model is based on several assumptions that may not be valid for the system investigated [2], [30] and [50]. Secondly, as is typical for the specific case

of water in macromolecular materials (but usual also in other heterogeneous systems) the 1H NMR spectrum can seldom be characterized in detail that would permit one to go beyond a simple two-state model. In other words, one can only distinguish between a narrow and a broad signal component, even if either or both are in reality composites of contributions from molecules with slightly different characteristics such as mobility. In other words, there might be a distribution of molecular properties such as exchange. In a two-state model, it is difficult to account for the effect of such distributions, particularly so if they are correlated (such as exchange rates and relaxation rates varying in parallel). Thirdly, as we already mentioned more than one magnetization exchange mechanisms might simultaneously be present, such as proton exchange and cross-relaxation for agarose.