Artificial intelligence has designed a radio pulse that achieves 99% accuracy in chemical analysis while using almost no power.
April 26, 2026
Original Paper
Artificial Intelligence-Designed Ultra-Low-Power Broadband Universal and Pulses for NMR Spectroscopy: Application to 19F NMR
SSRN · 6626947
The Takeaway
Nuclear Magnetic Resonance (NMR) spectroscopy is a vital tool for chemistry, but it often hits a wall because the hardware gets too hot to handle high-power pulses. This study used AI and evolutionary algorithms to design a specific type of universal pulse that stays ultra-cool. The AI-designed pulses achieve 99% fidelity across a wide bandwidth while requiring a tiny fraction of the power of traditional methods. This allows researchers to perform high-precision 19F NMR experiments that were previously impossible with cryogenic probes. It removes a major physical limitation on how we see the internal structure of molecules. This breakthrough will lead to cheaper, faster, and more sensitive medical imaging and chemical testing.
From the abstract
With the availability of high and ultra-high magnetic field strength, it becomes essential to develop innovative radiofrequency (RF) pulse shapes for NMR spectroscopy that can not only excite, but also efficiently invert and refocus nuclei with broad chemical-shift dispersions. Ideally, these new pulses should provide broad irradiation bandwidth, high operational fidelity, and superior inhomogeneity compensation. While classical hard pulses are inadequate to meet these requirements, broadband an