Diffusion-ordered NMR spectroscopy (DOSY) has been dure, of course, leads to confusing results when peaks overused in the analysis of a variety of complex mixtures, e.g., lap, but can be justified when overlap is rare. Furthermore, samples containing large numbers of monodisperse compothis method can resolve nonoverlapping discrete components nents (discrete mixtures) and polydisperse polymers (conin unknown samples when diffusion coefficients differ by tinuous mixtures) (1, 2). The basic requirements for DOSY less than 20%. Therefore, there is strong motivation for deare (a) an undistorted pulsed-field-gradient NMR (PFGveloping DOSY experiments with limited signal overlap in NMR) data set obtained with selected values of the gradient the NMR dimension(s). To this end, resolution has been pulse amplitude g and duration d, and (b) provisions for improved through stop-and-go sample spinning (9), eddydata inversion through an approximate inverse Laplace transcurrent suppression (10), use of high fields (1), and the use formation (ILT) or other method to obtain the ''most likely'' of 3D DOSY experiments based on 2D NMR (3). Continuspectrum of diffusion coefficients at each chemical shift. We ing this effort, we report in this Note a major resolution emphasize that the generation of a diffusion spectrum with enhancement by means of a novel heteronuclear-detected peaks and distributions plotted versus the diffusion coeffi-DOSY experiment incorporating coherence transfer (11). In cient is a necessary condition for DOSY. It is the spectral the following, we specialize the discussion to 13 C as repredisplay that permits distributions to be evaluated and multisentative of low-abundance heteronuclei with small gyrople components in polydisperse samples to be detected. Also, magnetic ratios. the peaks in 2D and 3D DOSY (3) displays permit molecular It is well known that 13 C NMR spectra provide better species to be easily distinguished and identified .
resolution than proton spectra because of wider chemical-In the data inversion, it is assumed the signal for the ith shift ranges, longer T 2 , and the virtual absence of homonudiscrete species is attentuated by the Stejskal-Tanner factor clear spin-spin coupling for low-abundant nuclei. Also, the exp[0D i (D 0 d/3)q 2 ], where D i is the diffusion coeffismaller gyromagnetic ratio of 13 C relative to 1 H means that cient for the ith species, D is the diffusion time, d is the spatial inhomogeneity of the main magnetic field B 0 and gradient pulse width, g is the gyromagnetic ratio, and the eddy currents have less effect on NMR linewidths and gradient-pulse area is q ร
ggd (5). There are two major, shapes. In addition, 13 C magnetization can be easily manipurelated problems in the data analysis for DOSY. The first lated by RF pulse sequences, e.g., INEPT or DEPT, to give is, of course, sensitivity, and the second arises from fundasimplified or sorted NMR spectra. Unfortunately, the low mental limitations on the amount of information that can be abundance, long relaxation times, and small gyromagnetic derived from diffusion data in the presence of noise. It is ratios also yield low sensitivity, a severe problem for DOSY well known that the resolution of a sum of exponentials with where high signal-to-noise ratios are essential for meaningful noise is an ill-posed problem (6). Therefore, advances in data analysis. Also, the low gyromagnetic ratio for 13 C reanalysis methods must rely on assumptions, prior knowledge duces the values of q that can be obtained in practical experiabout the sample, and the simultaneous use of information ments and presents a problem when low diffusion rates must at various chemical shifts in the FT-PFG-NMR data set be measured with commercially available modest gradients. . We expect that only incremental improvements However, the heteronuclear DOSY experiments with coherwill be made in the analysis area.
ence transfer presented here can improve sensitivity and It is our experience that highest accuracy is obtained in avoid the problem of low q values by transferring polarizasituations where a single exponential function can be fitted tion and diffusion information from protons to 13 C prior to to experimental data by nonlinear regression. This procedetection. Some typical pulse sequences for heteronuclear-detected DOSY with coherence transfer are shown in Fig. . The