Book X · Ch. 5 · 2026-09-17 · Custody
The Notebooks and What Reached Us
Ramanujan wrote down over a hundred class invariants and thirty singular moduli without a word of proof and without saying what they were. The recovery took years, and it was not a recovery of proofs. It was a recovery of meaning. This volume is about what survives transmission, and this is the cleanest case in mathematics.
Custody · what a number loses on the way to you
The usual account of Ramanujan is a loss of time — died at 32, imagine the theorems. The more interesting loss is of labels. The values arrived. What did not arrive was any statement of what they were values of, and reconstructing that took a generation.
1 · The values arrived without their names
Berndt, Chan and Zhang, who established the singular moduli, describe the problem in a sentence that belongs in this volume more than in a mathematics paper:
Berndt, Chan and Zhang, 1997
“In all cases, Ramanujan simply recorded in his notebooks the values or factors of singular moduli without explaining their meanings. It took us several years to discover that these radical expressions were singular moduli.”
Not years to prove them. Years to work out what they were of. The content survived the journey; the frame did not. A page of radicals with no statement of what function was being evaluated at what argument is a message whose payload is intact and whose header is gone.
2 · What the corpus did with one of them, unknowingly
Watson's algorithm, extracted from page 320 of the first notebook, produces at $n = 58$ the integers $U = 1$, $V = W = 9801$, $S = 9802$. Chapter R computed that row, verified it to thirty places, printed it, and used none of it.
9801 is $99^2$, and it is the prefactor of the most famous series Ramanujan ever wrote for $\pi$. The corpus was holding the number and not the label — the identical failure, one century downstream, performed by people with the whole literature available. That is the argument of this volume in miniature: custody is not storage. A value you cannot connect to anything is a value you do not have.
3 · Transmission has a direction, and the losses differ
| What moved | What survived | What was lost |
|---|
| Notebooks, Madras → Cambridge | the values, exactly | the meanings; recovered decades later at large cost |
| Letters to Hardy, 1913–14 | 120 theorems | provenance — some known, some new, some wrong, none marked |
| Watson's proof of the algorithm | certainty that it works | any account of how it was found; the authors looked for a better proof and did not find one |
| Bailey–Borwein–Borwein, scanned 2026 | the whole paper | one theorem statement, destroyed by OCR (§5) |
The third row is the one that unsettles a corpus built on verification. Watson's proof is a verification; it establishes the result and explains nothing. This series carries 122 verification scripts with the same property. They are the reason anything here can be checked without trusting the author, and not one of them has ever explained anything.
4 · A live transmission defect, in this corpus, unresolved OPEN
$k_{210}$ — open since the Ramanujan chapter was written
In his second letter to Hardy, Ramanujan asserts a value for $k_{210}$ as a product of eight powers of units. Transcribed from the scanned page exactly as printed and evaluated at 80 digits, that product equals $\alpha_{210}$, not $k_{210}$ — the square of the labelled quantity — agreeing to 78 significant figures while disagreeing with $\sqrt{\alpha_{210}}$ at relative error 1.
Two readings, and arithmetic cannot choose. Either the eight exponents were misread from the scan and each should be halved, or the expression is $\alpha_{210}$ carrying Ramanujan's own label rather than the modern one. What is not in question: the radicals are right and the value is exact. Only the exponent on it is open, and resolving it needs the printed page.
Note what kind of error this is. Not a wrong number — the number is right to 78 figures. A wrong name for it, which is precisely the failure mode of §1, still live, in a document produced this year.
5 · The same loss, at machine speed
The scan of Bailey–Borwein–Borwein used for the $1/\pi$ work has clean text throughout except the theorem defining the series parameter, where the OCR collapses: x2Nn;1, (1 + k2 )2. A page that survived 37 years in print was destroyed in the act of being made searchable.
What was done about it belongs in Vol XIII: the theorem was not transcribed. Four candidate readings were tested against an integer the series forces, and one closed exactly. The damaged text stayed damaged; the identity was recovered by arithmetic. That is a custody technique, not a mathematical one — a way of reading a document too damaged to read.
6 · The procedure this chapter emits
What to record so a number survives
- The label, in full. Not $\alpha$ but “$\alpha_n = k(e^{-\pi\sqrt n})^2$, the square of the modulus”. §1 and §4 are both failures of this line alone.
- The convention, named. $k$ or $k^2$ is a convention, and a century later nobody can tell which you meant from the value.
- The source, pinned. A commit hash, an edition, a page. wp107-verify.py records that it fetches bytes over HTTPS without pinning a digest — upstream can change its answer without changing it.
- The failure, kept. When a value disagrees with its label, record both and the disagreement, as §4 does. A defect deleted is a defect that recurs.
7 · Sources
- B. C. Berndt, H. H. Chan and L.-C. Zhang, Ramanujan's Singular Moduli, The Ramanujan Journal 1, 53–74 (1997).
- D. H. Bailey, J. M. Borwein, P. B. Borwein, Ramanujan, Modular Equations, and Approximations to Pi, Amer. Math. Monthly 96 (1989).
- Producing scripts: book7/ch-ramanujan-verify.py, book7/ch-ramanujan-1pi-verify.py.