Nature Physics · 12 November 2026 · From issue 66
Optimal learning of quantum channels in diamond distance
A method learns any quantum process using the fewest possible test runs, and no method can do better.
Antonio Anna Mele, Lennart Bittel.
Scientists often need to figure out exactly what an unknown quantum process does to particles, a task called process tomography. This paper finds the exact number of times you must run that process to learn it accurately, measured by a standard yardstick for comparing quantum processes. The authors built a method that hits this exact number and proved no method can beat it. They did this by turning the hard problem into an easier, well known one: learning a pure quantum state.
This bounds the best possible performance for quantum process tomography, shown only in theory.
Theoretical proof only, no hardware test, result sets a hard limit.
Integrity screen: passed (3 checks) Checked 19 November 2026. Retraction record: none. DOI resolves at doi.org. Metadata record found (Nature Physics). Read the source
This screen checks process signals. It does not tell you the findings are true.