The tiny difference between hard pi pulses and their delta-function approximation can be exploited to control coherence. Variants on the magic echo that work despite a large spread in resonance offsets are demonstrated using the zeroth- and first-ord
er average Hamiltonian terms, for 13-C NMR in C60. The 29-Si NMR linewidth of Silicon has been reduced by a factor of about 70,000 using this approach, which also has potential applications in magnetic resonance microscopy and imaging of solids.
In spectroscopy, it is conventional to treat pulses much stronger than the linewidth as delta-functions. In NMR, this assumption leads to the prediction that pi pulses do not refocus the dipolar coupling. However, NMR spin echo measurements in dipola
r solids defy these conventional expectations when more than one pi pulse is used. Observed effects include a long tail in the CPMG echo train for short delays between pi pulses, an even-odd asymmetry in the echo amplitudes for long delays, an unusual fingerprint pattern for intermediate delays, and a strong sensitivity to pi-pulse phase. Experiments that set limits on possible extrinsic causes for the phenomena are reported. We find that the action of the systems internal Hamiltonian during any real pulse is sufficient to cause the effects. Exact numerical calculations, combined with average Hamiltonian theory, identify novel terms that are sensitive to parameters such as pulse phase, dipolar coupling, and system size. Visualization of the entire density matrix shows a unique flow of quantum coherence from non-observable to observable channels when applying repeated pi pulses.
