Ten to the fourteen point five

This correction is the archive’s biggest miss: a renormalisation group cascade that claimed a physical regime change and missed the target by fourteen orders of magnitude in the exponent, over sixty orders in the probability. The post is short because the finding is simple, and it is here because the failure mode, an extrapolated exponent quoted as a prediction, is one of the most common in theoretical work and one of the least policed.

The claim

The framework’s early dynamical papers included a cascade mechanism: as the substrate’s degrees of freedom integrate out, the boundary’s effective description flows through a sequence of regimes, and the claim was that the flow terminates at a scale where the boundary’s physics undergoes a phase transition, the one the programme’s novelty story needs. The paper computed the flow’s trajectory through several orders and quoted the terminal scale.

The audit

The audit recomputed the flow and then did what flow analyses rarely receive: it summed the series. The cascade’s exponent, evaluated term by term and summed, comes to minus 139.3. The paper’s printed terminal exponent was minus 46, which the mechanism needed for its transition to land at the required scale.

The gap between minus 139.3 and minus 46 is not a correction. It is fourteen orders of magnitude in the exponent, sixty in the quantity. The printed number was not the sum of the series; it was the sum of the first few terms, eyeballed toward the value the mechanism required. The full sum, computed, walks the flow into a regime where the claimed transition simply does not occur: the cascade never terminates where the paper said, and the mechanism’s predicted behaviour is absent from the flow’s actual trajectory.

What was and was not salvageable

Nothing in the cascade’s structure was wrong: renormalisation group flows are real, the framework’s boundary theory has a genuine flow, and the computation’s early terms were correct. What failed was the inference from partial sum to terminal behaviour, and that failure is not repairable by adding terms: the full sum is what it is. The mechanism is marked open in the ledger, meaning the framework does not currently know where the boundary’s flow terminates or whether the needed transition exists. It is on the open problems list, with the corrected sum attached, and any future claim about the flow’s terminal behaviour must start from minus 139.3 rather than from the value the story wants.

The general lesson

The reason this failure mode deserves its own post is its respectability. Truncating a series and reporting the trend is standard practice in flow computations, and in most contexts it is honest, because the physicist’s judgement of trend is calibrated by experience. The failure happens at the boundary of calibration: exponents of order a hundred are not in any physicist’s calibrated range, and trend-eyeballing across fourteen orders is not extrapolation, it is invention. The audit protocol’s rule, generated by this corpse: any quoted series sum gets summed, mechanically, before any inference that depends on its value, and the mechanical sum is what gets printed. Judgement chooses the terms. Arithmetic chooses the total.

The next entry sets a rate equal to itself, and the two differ by ten to the fifty-eight: a rate that could not equal itself.