Evolution strategy optimization for selective pulses in NMR.

We present a first set of improved selective pulses, obtained with a numerical technique similar to the one proposed by Geen and Freeman. The novelty is essentially a robust and efficient "evolution strategy" which consistently leads, in a matter of minutes, to "solutions" better than those published so far. The other two ingredients are a "cost function," which includes contributions from peak and average radiofrequency power, and some understanding of the peculiar requirements of each type of pulse. For example, good solutions for self-refocusing pulses and "negative phase excitation pulses" (which yield a maximum signal well after the end of the pulse) are found, as may have been predicted, among amplitude modulated pulses with 270 degrees tip angles. Emphasis is given to the search for solutions with low RF power for selective excitation, saturation, and inversion pulses. Experimental verification of accuracy and power requirements of the pulses has been performed with a 4.7 T Sisco imager.