Work

A checked case classification of the E-function slice of continued fractions with linear partial numerators and denominators

Year: 2026
Type: preprint
Abstract: We classify the continued fractions b_0 + K(n>=1) (a_0 + a_1 n)/(b_0 + b_1 n) with rational coefficients, b_1 != 0 and no vanishing partial numerator. Each such fraction either converges to a value given by Kummer's function 1F1 or by 0F1, or has no finite limit in Perron's sense. The value is rational in exactly the cases listed in Theorem 1 and transcendental in all other convergent cases. The classification is assembled from Perron (1913), §81, and the Siegel–Shidlovskii theory in the form of Beukers (2006). The proofs are proofs modulo statements pinned in the cited sources. An attribution table records which statements we could and could not locate in the literature, and a census of 531,441 coefficient tuples serves as an illustration and a consistency check. No claim of priority is made. What is classical. The evaluation of the fractions as ratios of confluent hypergeometric series and the criterion for a finite limit follow from Perron's Die Lehre von den Kettenbrüchen (1913), §81, together with facts on Kummer's function from the DLMF. The transcendence statements follow from the Siegel–Shidlovskii theorem; they are proved here from Corollary 1.4 of Beukers (2006) with an irreducibility lemma, and they are expected to be classical. Siegel's 1929 memoir, Shidlovskii's book and Lorentzen–Waadeland (1992) were not consulted. The convergence statements rest on a stated reading of Perron's Satz 46, called R1 in the note. Checks. Two classifier programs, written from one classification specification, were sealed in a project ledger before their outputs were compared. They agree on all 531,441 census rows (coefficient tuples); this agreement shows that both programs read the specification the same way and is not independent evidence about the specification. Both reproduced the expected outputs of the 20 operator controls, which were fixed before the comparison. A separate numerical recheck evaluated all 4,608 in-slice rows. The sealing and the comparison are recorded as ledger seq 16 (entry_sha256 c1e8451eb5883d7b902ff24c0bf227ee39e923307c3e7ddfab2e0fc65bd598d5) and seq 17 (entry_sha256 505695accb3792fd3343812a1c7e212877a65486f96b4d03baec330f63952b37); the ledger is local and append-only. These are consistency checks, not proofs, and no part of the classification has been refereed outside the project. The source searches, computations, checks and drafting of this note were carried out with AI assistance (Claude by Anthropic, and AI agents run through the GitHub Copilot CLI), under a pre-registered and hash-ledgered protocol. The author directed the work, made every decision recorded here, and takes responsibility for the content. Files. note.pdf (the note), and the supplementary files: C12_RESULT.md, C12_result_tables.md, C12_final_classification.csv, controls_C12.csv, numeric_recheck_v2.py, numeric_recheck_results.csv, numeric_recheck_summary.txt, numeric_recheck_v2_run1.out.txt, prereg_numeric.md, pinned_statements.md. Licensing. All files of this record, including the script, are under CC BY 4.0.
Author: Papanokechi
Language: English
Open Access status: green
OpenAlex ID: W7236072538
Cites:
Cited by: 1
Related to: 10