Repetitive heteronuclear decoupling schemes are suscepti- spins interchanges theaand blabels. For I spins at reso-ble to the problem of cycling sidebands (1). Viewed in the nance near the center of the sweep, the adiabatic pulse causestime domain, decoupling is basically a repeated refocusing reconvergence after a time period of approximately 0.5T,of the divergence of magnetization vectors due to the spin– i.e., near the end of the sweep. They reach full divergencespin coupling, and unless the focusing is precise and the again near the center of the next adiabatic sweep, ready forsampling of the observed signal is exactly synchronized with refocusing. Consequently, these principal sidebands are mostthe focus points, spurious modulation is normally introduced intense for spins with chemical shifts near the middle of theinto the free induction decay. The Fourier transform of these sweep.oscillatory artifacts consists of pairs of modulation sidebands The subharmonic sidebands also have a phase gradientflanking the resonance frequency and separated from it by that is a function of the decoupler sweep, and which isthe cycling frequency. These cycling sidebands are undesir- canceled by alternating the sweep direction (Fig. 2). How-able because they can eventually interfere with the detection ever, these sidebands are most prominent near the edges ofof meaningful NMR signals, for example, the cross peaks the sweep. For a chemical shift near one edge, the corre-in two-dimensional NOESY spectra. In general, the cycling sponding vectors aand bdiverge for almost the entire sweepsideband problem is more serious at low decoupler power, duration T and reconverge to a focus near 2T, at which timewhich is the preferred operating condition. the next adiabatic sweep can have little effect. We must waitNow that efficient adiabatic decoupling schemes have until 3T for full divergence before spin inversion is oncebeen developed (2–14), it is possible to cover chemical- again effective. The resulting modulation frequency isshift ranges that far exceed even those required for carbon- 1/(2T) hertz.13 decoupling in a field of 18.8 T. Consequently, radiofre- The principal and subharmonic sidebands cannot be can-quency heating is no longer a serious problem, but some celed by simply desynchronizing the decoupling and sam-care must be taken to minimize cycling sidebands. We distin- pling operations; this only spreads the sidebands across theguish two main types— principal cycling sidebands that oc- entire decoupling bandwidth. [The asynchronous decouplingcur at frequency offsets of {1/(T) hertz, where T is the proposed for homonuclear systems (15) addresses a differentadiabatic pulse duration, and subharmonic sidebands which problem, i.e., the artifacts created by the presence of a modu-appear at {1/(2T) hertz. lated radiofrequency field (16).] Two different proceduresThe variation of the principal cycling sidebands as a func- are recommended for suppression of the two types of side-tion of the decoupler sweep (Fig. 1) shows a progressive bands. Subharmonic sidebands are greatly attenuated by en-change of phase through almost {1807 from one edge of suring that spins at all chemical shifts experience the adia-the sweep to the other, but if the sweep direction is reversed, batic spin inversion as nearly as possible at the same time.the dispersion contributions are inverted. Averaging over This suggests the use of shorter adiabatic sweeps inter-both sweep directions leaves these sidebands everywhere in spersed with ‘‘windows’’ of zero decoupler intensity, im-pure absorption (Fig. 1c). Clearly, the principal sidebands plying operation at a higher peak decoupler level, B