A 4–30-hour scan of the 73-hour January 2026 TESS block of 3I/ATLAS finds a nine-hour signal that depends on the core model

We test brightness cycles in the longest January 2026 TESS block of interstellar comet 3I/ATLAS. We use 1216 selected images from a 73.4-hour block. We fit a median coma image, image derivatives, a compact template, and sky terms. We scan 601 frequencies from 4 to 30 hours with eight image models and two time trends. The strongest fractional improvement occurs near 9.3 hours with a 1.3-pixel Gaussian template and a quadratic trend. It removes 22.2% of the baseline residual sum of squares. Joint block sign controls give tail fractions 0.0005, 0.0020, and 0.0155 for 1-, 2-, and 4-hour blocks. These values require independent blocks with sign-symmetric errors. The primary 1.0-pixel model gives a 0.492% semiamplitude near 9.4 hours and a 2-hour tail fraction of 0.526. With the wider template, a fit at 9.3 hours remains after an edge cut and has similar phase in both halves. Signed image maps contain opposite-sign core lobes. Artificial signals that match the primary template retain 98.64–99.67% of their input amplitude. A change of source width changes this response to 55–137%. The result is a model-dependent image signal. Its cause is unresolved. These tests do not establish a nucleus rotation period or a physical upper limit. The package includes source pixels, code, hashes, results, and reproduction instructions.

3I/ATLAS · TESS · interstellar comet · time series · image subtraction · signal injection · correlated noise · reproducible research