MAGNETIC RESONANCE IN CHEMISTRY Magn. Reson. Chem. 36, 706È714 (1998) Pulse sequences for high-resolution diþusion-ordered spectroscopy (HR-DOSY) Ò Michelle D. Pelta,1 Herve  Barjat,1 ¤ Gareth A. Morris,1 * Adrian L. Davis2 ” and Stephen J. Hammond2 1 Department of Chemistry, University of Manchester, Oxford Road, Manchester M13 9PL, UK 2 Shell Research Limited, SRTCT, P.O. Box 1, Chester CH1 3SH, UK Received 9 March 1998 ; revised 1 May 1998 ; accepted 1 May 1998 ABSTRACT : Six pulse sequences are described, all based on the stimulated echo, for use in high resolution di†usion-ordered spectroscopy (HR-DOSY). HR-DOSY requires spectra with clean baselines, pure phases and lineshapes that are independent of Ðeld gradient pulse amplitude. Lineshape problems arising from the static Ðeld perturbations caused by Ðeld gradient pulses and phase errors caused by zero quantum coherence in strongly coupled spin systems are discussed, and the performance of the six sequences is compared. Pulse sequences which use balanced pairs of antiphase Ðeld gradient pulses show signiÐcant advantages. 1998 John Wiley & Sons Ltd. ( KEYWORDS : NMR ; DOSY ; stimulated echo ; bipolar pulse sequences ; zero quantum coherence INTRODUCTION Di†usion-ordered spectroscopy (DOSY)1h11 seeks to separate the NMR signals of di†erent species according to their di†usion coefficients. A series of spin-echo spectra are measured with di†erent pulsed Ðeld gradient strengths, and the signal decays are analysed to extract a set of di†usion coefficients with which to synthesize the di†usion domain of a DOSY spectrum. In 2D DOSY the initial di†usion-weighted spectra are one- dimensional ; adding di†usion weighting to 2D experi- ments such as COSY, NOESY or HMQC gives 3D DOSY12h17 spectra. DOSY can be roughly subdivided into two areas : low-resolution DOSY,1h7 of simple mix- tures of molecules with widely di†ering sizes and poorly resolved NMR spectra, and high-resolution DOSY (HR-DOSY),8h11 of more complex mixtures which may contain molecules of very similar sizes but which give well resolved NMR spectra. The difficulty of extracting clean di†usion parameters from overlapping signal decays sets strict limits on the di†usion resolution of low-resolution DOSY, but in the high-resolution case relatively small di†erences (of the order of 1%) in di†u- sion coefficient can be resolved. To identify such small di†erences reliably it is necessary that the systematic errors be reduced to an absolute minimum : the chal- lenge of designing successful HR-DOSY techniques is to * Correspondence to : G. A. Morris, Department of Chemistry, Uni- versity of Manchester, Oxford Road, Manchester M13 9PL, UK E-mail : g.a.morris=man.ac.uk ¤ Present address : University Chemical Laboratory, LensÐeld Road, Cambridge CB2 1EW, UK. ” Present address : PÐzer Central Research, Sandwich, Kent CT13 9NJ, UK. Ò Dedicated to Professor John D. Roberts on the occasion of his 80th birthday. Contract/grant sponsor : EPSRC ; Contract/grant number : GR/ K16296 ; Contract/grant number : GR/K44619 ; Contract/grant number : GR/L 17443. combine experimental methods which minimize such errors with data analysis procedures which compensate as accurately as possible for the errors that remain. This paper is primarily concerned with evaluating the impact of systematic errors with six di†erent pulse sequences, three of which have been described previously and three of which have not. In HR-DOSY the simplifying assumption is generally made that each peak in the NMR spectrum corresponds to a single species ; Ðtting the decay of the height of each peak as a function of the square of Ðeld gradient pulse area to a single exponential then yields a di†usion coef- Ðcient and a standard error for each signal in the NMR spectrum. The di†usion dimension of the DOSY spec- trum is synthesized by placing a Gaussian signal in the di†usion domain for each peak in the NMR spectrum. The Gaussian is centred on the Ðtted di†usion coeffi- cient, has a width proportional to the standard error obtained from the Ðtting process, and is of volume or amplitude proportional to the initial NMR peak height. The best relative accuracy in di†usion coefficient is obtained if a two-parameter (initial amplitude and decay constant) Ðt is used, but this requires that all signals decay towards zero, and therefore makes main- tenance of a clean spectral baseline important. The e†ect of any baseline o†sets is to bias the decay con- stants and to degrade the standard errors obtained. Care should therefore be taken to use experimental con- ditions (e.g. acquisition timing, choice of analogue Ðlter bandwidth) that give Ñat baselines, and after initial Fourier transformation it may be helpful to apply base- line correction. It is also important to ensure that all signals have pure absorption phase. Even small phase errors can have signiÐcant e†ects, since the broad tails of dispersion mode signals extend far further than the skirts of an absorption mode lineshape. The dispersion tails have two undesirable e†ects : they greatly increase ( 1998 John Wiley & Sons, Ltd. CCC 0749-1581/98/100706È09 $17.50