An inverse-frequency width rule reduces source-balanced error by 1.4 percent in 223 selected CHIME events

We test whether the first and last fitted components of the same radio burst have a constant product of width and peak frequency. We use CHIME/FRB Catalog 2, release 20251231. In 223 selected events from 120 catalog sources, an inverse-frequency rule reduces squared log-width-ratio error by 1.4 percent relative to equal widths. Equal source weights and 10000 source resamples give a 95 percent interval from -3.4 to +5.7 percent. The fitted log-ratio slope is -0.90, with an interval from -2.29 to +0.39. Thus, the data do not show a clear error reduction and do not exclude a broad inverse trend. At fixed fitted peak frequencies, 117 endpoint pairs cannot satisfy the exact rule within three quoted width errors. Allowing each peak to move by 20 MHz leaves 108 such pairs. These are conditional interval checks, not significance tests. Resolution, band-edge, scattering, pulse-separation and source controls show no clear error reduction. Missing peak errors, shared fit errors, component selection, and limited time resolution prevent a physical exclusion. No new source, planet, or emission mechanism is confirmed. Code, frozen input rows, provenance, results and offline steps are included.

astronomy · CHIME · fast radio bursts · component width · frequency scaling · conditional model test · null result