N-type COSY spectrum, obtained on a Varian Unity 500 The sensitivity of many 2D experiments is limited not by spectrometer, of the tetracyclic orthoamide: the ratio of signal to thermal noise, but by the ratio of signal to t 1 noise. This noise, in bands running parallel to f 1 at the f 2 frequency of strong signals, arises from instabilities in the radiofrequency phase, gain, field/frequency ratio, etc., of the spectrometer (1). These lead to pseudorandom modulations of the signals as a function of t 1 , which on Fourier transformation give rise to the familiar bands of ''noise.'' In this Communication, it is shown that the combined use of gradient-enhanced techniques (2) and reference deconvolution (3-5) can lead to a dramatic reduction in t 1 noise in 2D data sets, and can lower the threshold of detectability of cross peaks in 2D experiments by more than an order of Bands of t 1 noise are clearly visible above the thermal noise, magnitude.
particularly at the f 2 frequencies of the reference, TMS, and Most NMR experiments rely upon the selection of signals of the residual protons in the solvent, acetone-d 6 . that have followed particular coherence-transfer path-It has been shown (2) that pulsed field gradients may be way(s), and the rejection of those that have followed other used instead of phase cycling to select particular coherence pathways (6). This has traditionally been done by phase pathways, and gradient-enhanced experiments are becoming cycling, in which the phases of pulses and receiver are varied increasingly popular, replacing their phase-cycled counterto give constructive interference of the signals of interest, parts. It is a common misconception that gradient-enhanced and destructive interference of unwanted signals. This is in experiments do not suffer from t 1 noise. Although it is true essence difference spectroscopy, and thus any change in that they generally exhibit lower levels of t 1 noise than their amplitude or phase of the signals recorded in different repetiphase-cycled analogue, all gradient-enhanced experiments tions will result in incomplete subtraction, and lead to artishow t 1 noise, and some common experiments show little facts such as t 1 noise in 2D spectra. In experiments such advantage over their phase-cycled progenitors in this particuas indirectly detected heteronuclear correlation experiments, lar respect. The reduction in t 1 noise basically occurs because e.g., HMQC or HMBC, the signals that must be suppressed in a gradient-enhanced experiment the only signals detected are much stronger than the required responses, and artifacts are those that have followed the selected coherence-transfer from incomplete suppression of these unwanted signals are pathway(s). Thus the t 1 noise observed in a gradient-enoften significantly larger than the thermal noise. The factor hanced experiment arises from unwanted perturbations of limiting the detection of small signals is then the ratio of the selected signals only, whereas spectrometer instability signal to t 1 noise, rather than the signal to thermal noise. during the course of a phase-cycled experiment can result Even in cases in which the wanted and unwanted signals in t 1 -noise contributions from all possible coherence-transfer are of a similar magnitude, such as homonuclear COSY or pathways. The reduction in t 1 noise for a gradient-enhanced NOESY, the amount of information that may be obtained experiment will thus be greatest for experiments such as from a spectrum is often still limited by t 1 noise rather than HMBC or HMQC, in which the signals that are not required thermal noise. Figure 1 shows a phase-cycled absolute value are much larger than those that are.
Figure 2 shows an N-type gCOSY spectrum of the same tetracyclic orthoamide sample that was used to acquire the data shown in Fig. 1, recorded under identical experimental