2s/2p energy split in the hydrogen emission spectrum, not predicted by the Dirac equation, but explained by quantum electrodynamics, which is one of the first great triumphs of that theory.
Note that for atoms with multiple electrons, 2s/2p shifts are expected: Why does 2s have less energy than 1s if they have the same principal quantum number?. The surprise was observing that on hydrogen which only has one electron.
Initial experiment: Lamb-Retherford experiment.
On the return from the train from the Shelter Island Conference in New York, Hans Bethe managed to do a non-relativistic calculation of the Lamb shift. He then published as The Electromagnetic Shift of Energy Levels by Hans Bethe (1947) which is still paywalled as of 2021, fuck me: journals.aps.org/pr/abstract/10.1103/PhysRev.72.339 by Physical Review.
The Electromagnetic Shift of Energy Levels Freeman Dyson (1948) published on Physical Review is apparently a relativistic analysis of the same: journals.aps.org/pr/abstract/10.1103/PhysRev.73.617 also paywalled as of 2021.
TODO how do the infinities show up, and how did people solve them?
www.mdpi.com/2624-8174/2/2/8/pdf History and Some Aspects of the Lamb Shift by G. Jordan Maclay (2019)
Made huge advances in radar.
Notably, Isidor Isaac Rabi was a leading figure there, and later he was head at the Columbia University laboratory that carried out the crucial Lamb-Retherford experiment and the anomalous magnetic dipole moment of the electron published at The Magnetic Moment of the Electron by Kusch and Foley (1948) using related techniques.
E.g. list of papers by Isidor Isaac Rabi which includes A New Method of Measuring Nuclear Magnetic Moment.
But unfortunately they don't have paper to paper citations.
neurotree.org/neurotree/faq.php explains that you have to contact an admin to download the database, kind of sad:
How can I export tree data for my own analysis?Registered users should contact the site administrator (admin at neurotree dot org) for instructions on how to export data from the tree database.
That page also explains how they disambiguate authors with the same name:
How do you identify researchers' publications?Publications data are drawn from two databases: Medline and Scopus. Because of the large number of researchers with the same name, a disambiguation algorithm is required to accurately link researchers to papers they have authored. We match authors to papers using a two-step process. First, we identify candidate publications based on a simple string match between researcher name and the author list. Second, we look for overlap between co-authors and other individuals in the researcher's mentor network (trainees, mentors, collaborators, etc), and label publications with overlap as high-probability matches. Thus a complete family tree is likely to produce more accurate publication matches.